Full text of "Proceedings of the International Patent Office Workshop on Information Retrieval / sponsored by U.S. Patent Office, Patent Office Society and supported by the National Science Foundation"

iV W

?e oosY"-

?e n^ 1

Onwe^

povV-'

ANNIVERSARY OF
THE UNITED STATES
PATENT ACT OF 1836

Digitized by the Internet Archive

in 2012 with funding from

LYRASIS Members and Sloan Foundation

http://archive.org/details/proceedingsofOOinte

PART II

Proceedings of the

International Patent Office

Workshop on Information

Retrieval

Sponsored by

U.S. Patent Office

Patent Office Society

and

Supported by the

National Science Foundation

U.S. DEPARTMENT OF COMMERCE
Washington 25, D.C.

For sale by the Superintendent of Documents, U.S. Government Printing Office. Washington 25, D.C. - Price .<1.25-

Letter of Transmittal

Washington, June 15, 1962

The Honorable The Secretary of Commerce,
Washington 25, D.C.

Dear Mr. Secretary :

I have the honor of transmitting to you the record of the proceedings observ-
ing the Celebration of the American Patent System commemorating the 125th
Anniversary of the Patent Act of 1836 in accordance with a Joint Resolution of
Congress and a Proclamation of the President of the United States. This record
includes Part I, the proceedings of a series of seminars on the American Patent
System, a report of an industrial exhibit, luncheons, the Anniversary Banquet
and Part II, the proceedings of a conference entitled The International Patent
Office Workshop on Information Retrieval.
Sincerely,

^CT^^*-^- ^-<K-j^y

Commissioner

Joint Resolution

Whereas there occurred on July 4, 1961, the one hundred and twenty-fifth anni-
versary of the enactment of the Patent Act of 1836 which created the pres-
ent examination system for the grant of patents and which established the
United States Patent Office as a separate and distinct Bureau; and
Whereas there will be granted in the year 1961 the three millionth patent since

the enactment of this Patent Act; and
Whereas the patent grant is a traditional incentive for the promotion of the prog-
ress of the useful arts thereby contributing notably to the well-being of people
everywhere; and
Whereas encouragement of invention is essential to the continued economic and
technological development of this Nation, particularly in the light of our
international relationships and obligations; and
Whereas it is fitting that this anniversary of this important Patent Act and the
establishment of the Patent Office which proved so advantageous to this
country and to the other countries of the world should be observed: Now,
therefore, be it

Resolved by the Senate and House of Representatives of the United States
of America in Congress assembled, That the Secretary of Commerce and the
Commissioner of Patents and such other persons or groups as they may designate
be authorized to make suitable arrangements for an appropriate observance to
show the important role of the American patent system and the United States
Patent Office in the growth and progress of the United States of America; and
be it further

Resolved, That the President of the United States be requested to designate
the week of October 15, 1961, as "The American Patent System Week", to
invite a general public commemoration of an event which has proved to be so
important to this Nation and to the world.
Approved September 22, 1961.

AMERICAN PATENT SYSTEM WEEK

By the President of the
United States of America

A PROCLAMATION

WHEREAS the year 1961 marks the one hundred and twenty-fifth anni-
versary of the enactment of the Patent Act of 1836 which created the present
examination system for the grant of patents; and

WHEREAS in the year 1961 there will be granted the three millionth patent
since enactment of the Patent Act; and

WHEREAS the grant of a patent is a traditional incentive for the promotion
of the useful arts and thereby contributes notably to the well-being of peoplr
everywhere; and

WHEREAS encouragement of invention is essential to the continued economic
and technological development of this Nation; and

WHEREAS it is fitting that the anniversary of the Patent Act and the estab-
lishment of the United States Patent Office should be observed; and

WHEREAS a joint resolution of the Congress approved September 22, 1961,
authorizes the Secretary of Commerce, the Commissioner of Patents, and such
other persons or groups as they may designate to make suitable arrangements
for an appropriate observance of the American patent system and the United
States Patent Office; and

WHEREAS the said joint resolution requests the President of the United
States to designate the week of October 15, 1961, as "The American Patent
System Week":

NOW, THEREFORE, I, JOHN F. KENNEDY, President of the United
States of America, do hereby designate the week of October 15, 1961, as the
American Patent System Week; and I invite the people during that week to
commemorate the American patent system which, by affording protection and
encouragement to inventors as envisaged and authorized by the Constitution,
contributes so greatly to the encouragement of inventive genius.

IN WITNESS WHEREOF, I have hereunto set my hand and caused the
Seal of the United States of America to be affixed.

DONE at the City of Washington this twenty-second day of September in
the year of our Lord nineteen hundred and sixty-one, and of the Independence
of the United States of America the one hundred and eighty-sixth.

By the President:

Acting Secretary of State ( fy /O

Contents — Part II

Monday morning, October 23, 1961
Opening Remarks — General Chairman

Honorable Horace B. Fay, Jr., Assistant Commissioner of Patents, U.S. Page

Patent Office 81

Address

Honorable David L. Ladd, Commissioner of Patents, U.S. Patent Office . . 81

Address

Honorable Robert C. Watson, Former Commissioner of Patents, U.S.

Patent Office 84

Remarks

I. J. Rotkin, Patent Classifier, U.S. Patent Office 85

D. P. Stein, Office of Research and Development, U.S. Patent Office . .
Discussion 88

Monday afternoon — October 23, 1961

Moderator — John C. Green, Director, Office of Technical Services 91

"The International Patent Classification System," Honorable Werner E.

Rubach, Director, Technical Section, German Patent Office 91

"The United States Patent Office Classification System" H. S. Vincent,

Patent Classifier, U.S. Patent Office 95

"Report of United Kingdom Study Group on Classification" Honorable

L. F. W. Knight, Superintending Examiner, Patent Office, United

Kingdom 101

Discussion 107

Tuesday morning, October 24, 1961
Moderator — Harold Pfeffer, Office of Research and Development, U.S. Pat-
ent Office 116

"A Suggested Interim Solution to the Patent Examination Problem,"
Dr. C.J. de Haan, President, The Netherlands Patent Office .... 116

Discussion 119

Moderator — Dr. Watson Davis, Director, Science Service 123

"Encoding and Abstracting for Universal Use," Honorable Cecil N.
Jeffery, Supervising Examiner (Classif.) Patent Office, Australia ... 123

Discussion 125

"Abstracting of Patents," Honorable Kotaro Otani, Chief Examiner,

Patent Office, Japan 126

Discussion 127

"Some Thoughts on Automatic Indexing and Encoding," Honorable Tore

Oredsson, Bureau Director, Patent Office, Sweden 128

Discussion 131

Tuesday afternoon, October 24, 1961
"Information Display for Encoding and Post-Retrieval Benefits," M.
Wolk, Supervisory Patent Examiner, U.S. Patent Office 133

Page

Discussion 153

"Problems in Analysis of Patents," H. Pfeffer, Office of Research and De-
velopment, U.S. Patent Office 156

Discussion 160

Wednesday morning, October 25, 1961
Moderator — Mr. S. N. Alexander, Chief, Data Processing Systems Division,

National Bureau of Standards 167

Problems in Mechanization of Search Procedures — "A Look at Patent Sys-
tems in Operation"

Cecil N. Jeffery, Australia 168

Werner Rubach, Federal Republic of Germany 169

P. van Waasbergen, International Patents Institute 172

L. F. W. Knight, United Kingdom 174

H. R. Roller, United States 176

Discussion 178

Moderator — Mr. Ezra Glaser, Section Chief, Data Processing Systems Divi-
sion, National Bureau of Standards 1 84

"Information Technology and the Law"
Reed C. Lawlor, Chairman, Electronic Data Retrieval Committee,

American Bar Association 184

Discussion 190

Wednesday afternoon, October 25, 1961

Remarks

Dr. C. J. de Haan, President, Patent Office, The Netherlands 193

Discussion — (Re: The Ad Hoc Committee) 196

"Theoretical Research on Information Processing Potentially Related to
The Information Retrieval Problem," Russell A. Kirsch, Data Process-
ing Systems Division, National Bureau of Standards 198

"On International Collaboration for a Complete Information Storage and
Retrieval System," Dr. Laurence B. Heilprin, Council on Library Re-
sources, Inc 202

"Mechanical Reading of Characters and Pattern Recognition" Mary
Stevens, Data Processing Systems Division, National Bureau of Stand-
ards 205

Discussion 208

Appendix A — "Planning a Long-Range Program for Mechanized Patent
Searching."

Martin Kalikow, General Electric Company 211

Appendix AA — "The Patent System and the Technological Revolution," Dr.
Alan T. Waterman, Director, National Science Founda-
tion 217

Appendix B — Summary of Answers to Questionnaire 225

Appendix C — Attendance List 234

Appendix D — Committees 239

Opening Session

Monday Morning, October 23, 1961

The opening session was called to order at
10 o'clock in a conference room of the De-
partment of State, Washington 25, D.C., by
the Honorable Horace B. Fay, Jr., Assistant
Commissioner of Patents, U.S. Patent Of-
fice, General Chairman.

Chairman Fay: Gentlemen, it is a pleasure to
welcome you all here for this International Patent
Office Workshop on Information Retrieval. I am
Horace Fay, Assistant Commissioner of Patents,
U.S. Patent Office.

The purpose of this meeting is, as you know, to
consider means of achieving cooperative efforts in
attacking and solving mutual problems of informa-
tion retrieval. Informal discussions, which have
been held with members of the staff of some of the
patent offices throughout the world over a period
of years, have clearly pointed up the community of
interest shared by these organizations in the prob-
lems of storage and retrieval of technological in-
formation. Even where the several patent laws
are superficially different from each other, the prob-
lems raised by the requirements for making a search
of the prior art are remarkably similar.

Although the state of the development of data
processing in several countries which have examin-
ing systems is not uniform, it is believed that many
of these countries can contribute skills and man-
power to an international effort. Apart from the
common interest in all aspects of information re-
trieval problems, which the patent offices share
with other users of information, the need for deep
analysis of the information content of documents,
the need for generic searching and the frequent
searching for equivalents, lead to specialized prob-
lem areas, which may be susceptible to a joint
research attack by the several patent offices.

The magnitude of the problems to be solved is
such that no single patent office is likely to be fully
successful in their solution. However, cooperative
work appears to offer a reasonable chance for

achieving success. The celebration of the 125th
anniversary of the United States Patent Office last
week, here, offered an opportunity for convening
representatives of the working staffs of a number
of patent offices which perform searching as one
step in the patent examining process. I think this
will outline for us the scope of effort we hope to
have during the next few days.

(At Mr. Fay's request, Mr. Gareau, President
of the Patent Office Society introduced the visitors
from abroad to the assembly. )

Chairman Fay: It is now my pleasure to introduce
to you the United States Commissioner of Patents,
David Ladd, who will begin our conference this
morning. Mr. Ladd.

Mr. Ladd: Gentlemen, it is a great pleasure to wel-
come you to the Information Retrieval Workshop,
and we are grateful to you for coming. I hope, in
these brief opening remarks to indicate the import-
ance and gravity that we attach to these meetings.
Before I explain generally the problem which has
led us to organize this Workshop on Information
Retrieval, I do want to express our appreciation to
the National Science Foundation which, by its
grant, has made this meeting possible, and to thank
also Mr. Pfeffcr, Mr. Gorecki, Commissioner Wat-
son and others for their efforts in organizing and
preparing this meeting.

The importance of this meeting, in m\ view, lie-
simply in the fact that the burdens of maintaining
an examining system, such as- we have in the United
States, are steadily becoming heavier. There are
several reasons for this. One is. the growing com-
plexity of the applications which arc filed. An-

other is the problem of personnel turnover, which
is not unknown to some of your patent offices. The
third, of course, is what we call the growth of the
cumulative search load. Simply it is the growth in
that material which must be studied and must be
searched to determine whether or not an application
discloses patentable and novel subject matter.

The program of research in information retrieval
in the Patent Office is, of course, directed to achiev-
ing to some degree at least, mechanical searching,
to replace with a mechanical search the work which
has traditionally been done manually. If we are
eventually to achieve general mechanized searching
in the Patent Office, this will not only improve our
chances of maintaining an examining system and
avoiding having it break down under us and com-
pelling us to go to other alternatives, but it should
also greatly strengthen the patents which do issue.

As most of you know, by virtue of the examination
accorded a patent application in the United States
Patent Office, it issues with a presumption of valid-
ity. General mechanized searching should also re-
duce instances in which the Patent Office does not
find relevant prior art. It should enhance the legal
presumption of validity and consequently it should
reinforce public and judicial confidence in the pat-
ent system.

In March of this year, a committee, under Dr.
Gilbert King, Director of Research of IBM, filed
with the Patent Office a report on the prospects of
automation in the search procedures of the Patent
Office. This report was a sequel and a supplement
to the now famous Kelly report which evaluated the
functions and operations of the Department of
Commerce, including the Patent Office in relation to
national scientific needs. That report was prepared
by a committee headed by Dr. Mervin Kelly, con-
sultant to the President of IBM. I have, inci-
dentally, brought with me this morning copies of
the King report in limited numbers, and they are
available here at the head table. If additional
copies are needed, we can have them reproduced.

Since the time that the King report was filed, I
asked Dr. Allen Astin, Director of the Bureau of
Standards, and Mr. John Green, Director of the
Office of Technical Services, to conduct another
study on our research and development program in
the Patent Office and to file their recommendations.
Both of these studies have now been made public.
This quotation from the King Committee report is
directly relevant to the point that I want to make
this morning. I am now quoting: "The whole

philosophy of the patent system is dependent on the
accuracy of the patent examiner's work. Some er-
rors are inevitable ; this is no reflection on the Patent
Office or its examiners, nor is our recommendation
that the error rate be measured. As time goes on,
it is quite likely that the rate will go up because the
search problem is becoming intrinsically more diffi-
cult, although the introduction of information re-
trieval systems may reverse this trend. If the error
rate increases sufficiently and it proves to be impos-
sible to reduce it, then our present patent system will
have to be abandoned or seriously modified. The
error rate may thus be regarded as important an
indicator of the health of the patent system as the
size of the backlog and the mean time of application
processing."

I can summarize the recommendations and find-
ings of both of these committees briefly, and they are
these: Operating under considerable handicaps in
a new field, the Patent Office research and develop-
ment staff, modest as it is, has achieved as brilliantly
as any comparable group of workers in the field.
However, the Patent Office problem of data proc-
essing is as severe as any existing. The work in this
field in the Patent Office is of importance to the
entire scientific community. It does not presently
lie within our capacity to conduct the exploratory
research necessary to achieve any longrun and gen-
eral solution. A judgment must be made as to
whether the Patent Office is willing to carry a dis-
proportionate burden of the research in this field.
There are strong reasons why it should. It starts
with the most sophisticated system of classification of
technical materials available. It is faced with its
own immediate and pressing problem, the solution
of which is, in fact, indispensable to the survival of
the examining system as we have known it. It has a
highly qualified cadre of technical people. And
finally, it represents an unparalleled opportunity for
costing out new systems as developed by comparing
the efficiency of manual searching and machine
searching.

Let me emphasize that the solution is not in sight
and we are very well aware of it. And I can empha-
size this point further by quoting to you from Dr.
Astin and Mr. Green's letter to me of July 14, 1961.
And again, I quote: "Confirming a point stressed
in the earlier reports that barriers to mechanization
lie in theory rather than in technology, we must
emphasize that many years of concentrated research
effort will be necessary before a significant impact
on the Patent Office search problem can be achieved.

Consequently, it would be unrealistic to expect ma-
jor reductions in the Patent Office backlog from
mechanization within the next few years."

It is my judgment that the Patent Office must
accept this challenge. It must do so, for the very
survival of the examining system as we have it is at
stake, and because of the unique contribution which
the Patent Office can make to the scientific com-
munity and to the country and to the world. For
fiscal year 1961, the American Patent Office has
budgeted a little over $500,000 for research and
development work. I think that you will agree
that this is a modest sum. We actually spent only
$460,000 of that appropriation. For this fiscal
year we have budgeted $580,000. The Astin-
Green report, which I have just quoted to you,
envisages an annual program of about $3 million.
That would be about 12% of our entire budget for
this fiscal year, which itself is at an all time high.
Of course, I must explain to you, as I am sure you
understand, that before the recommendation of the
Astin-Green report can become reality, those appro-
priations must be approved by the Department of
Commerce, the Bureau of the Budget and the
Congress.

The Astin-Green report also recommends the
appointment of a third assistant commissioner
whose sole responsibility will be the direction of
research and planning. We have now located that
man and his appointment will be announced within
the next few weeks.

While this problem of information retrieval and
mechanized searching is our greatest challenge, it
is also a splendid opportunity, for it means that the
Patent Office of the future will promote innovation
not only by securing rights in industrial property by
the granting of patents, but will also become in-
creasingly a center for the dissemination of technical
and scientific information. And on this point
again, I wish to quote at some length from the King
report.

"The problem (of information retrieval) is, of
course, a general one throughout science and tech-
nology and particularly worthy of attention for that
reason. In the seventeenth century there were men
who kept abreast of all science, but today even the
fragments into which all subjects have been split are
vast, and a scientist is hard put to keep up with the
research published on a small segment of a single
field. In spite of the growing number of informa-
tion services, information gaps become more preva-
lent and more difficult to fill. Mathematicians have

been known to work out a theory from scratch
rather than attempt to find it in the literature; a
chemist will devise a new synthesis rather than
search the journals; a physicist may perform an ex-
pensive experiment rather than look for previous
work on the subject. It could very well be true
that more than half the effort of the scientific and
technological personnel of the world is now being
spent in a duplication of previous effort. If this
estimate is correct, then an efficient method of
making knowledge available would double the
effectiveness of the scientific community.

"It is interesting to note, in returning to the Patent
Office problem that its relationship to the general
problem of access to scientific and technical in-
formation has a historical basis. One of the pri-
mary purposes of the patent system, as originally
devised, was to serve as an information clearing
house. Unfortunately, this latter function has all
but disappeared." There might be some who
would take some exception to the last remark.

"That steps be taken towards mechanization is
dictated, of course, by the sheer quantity of the
material to be searched. There is little question,
however, that the effectiveness of very high-speed,
mechanized searching procedures will depend on
the formulation of adequate searching and retrieval
theory. The Patent Office has quite properly set
up an Office of Research and Development to study
the possibilities of obtaining mechanical or elec-
tronic assistance. This office has a number of
excellent research contributions to the general prob-
lem of retrieval and has started a number of small-
scale projects to help the searchers in particular
areas. One of these projects, the steroid chemistry
file, has already proved valuable not only within
the Patent Office but also to the drug industry as a
whole. It is well worth noting that one of the
reasons for the success of this project is its contribu-
tion to the forgotten clearing-house function of the
Patent Office alluded to above."

These remarks, and especially these quotations,
I think, underline the importance which we attach
to these meetings. We welcome your cooperation.
This obviously will not be a short-run job. It will
last for years. But we must get started and we must
expand our work. At least, the American Patent
Office must expand its work strongly in this field
in the years ahead.

It is now my pleasure to introduce to you the
man under whom the Patent Office commenced its
work in this field. He is well-known to all of you

and no elaborate introduction is necessary. It gives
me great pleasure to introduce to you Robert C.
Watson, the former Commissioner of Patents.

Mr. Watson: Mr. Commissioner, Mr. Assistant
Commissioner, guests from countries other than our
own, members of the Patent Office staff and others.
It gave me great pleasure to hear your Commissioner
speak as he did, giving us all assurance that the vital
study which has been undertaken will be continued
and augmented. It is not exactly factual, however,
to say that I initiated the study. I had a predecessor
who initiated a study of similar import. Unfor-
tunately, however, he decided that the effort was not
worthwhile and discontinued it. Perhaps my early
interest was stimulated, in fact I am sure it was, by
a suggestion which appeared in a printed report of
the Senate Committee on Appropriations in 1954.
I read that report one day and established a com-
mittee the following day. The Commerce Depart-
ment also took prompt action, Dr. Vannevar Bush
being appointed head of a committee to consider
the search problem of the Patent Office and to make
recommendations. That committee functioned for
a number of months and rendered a report, copies of
which, I suppose, are still available for those who are
interested. The report recommended strongly that
the Patent Office, in conjunction with the Bureau of
Standards, as the Commissioner has pointed out,
should proceed vigorously to do various things.

One was to speed the classification effort so as to
bring about the earlier classification of the large
backlog of unclassified patents in our Office. And
the second one was to develop a system to search
the chemical patent literature as soon as possible.
That, and other recommendations, we implemented
as best we could. As the Commissioner has pointed
out, very considerable progress has already been
made which is of value not only to the Patent
Office, but to the whole of industry.

You gentlemen are all here because you know
that the success of the effort to mechanize patent
searching must succeed if we are to maintain our
examination systems. Perhaps I can say that, as
one of the older lawyers engaged in patent practice,
and who came into that practice through the search-
room door, that we have in this country, and there
exists in every examining country, a system which, to
my way of thinking, may be appropriately designated
a manufacturer's patent system. We issue patents
only after the applications therefore have been most
carefully scrutinized so that the manufacturer who
studies them knows exactly what they mean.

I imagine that everyone of you has visited our
search room and seen there engaged in inspecting
patents, on an average day, between three and
four hundred people. The search room is wholly
inadequate. We do not have enough seats. Re-
peated polls of those who are engaged in searching
in that room have disclosed the fact that a very
considerable proportion of them — as I recollect it,
somewhere between 50% and 60% of those search-
ing — were either studying the claims of unexpired
patents for the purpose of ascertaining the scope of
those claims or were trying to find anticipations of
those claims. In other words, they were engaged
either in infringement searches or in validity
searches, both of which make necessary the study
of the precise wording of the claims. As a result of
these studies, advice is communicated to the manu-
facturer, or the prospective manufacturer. It is
thus clear that the accurate and painstaking proc-
essing of those claims is essential to the success of
our system. That is the reason why we spend so
much time in the examining divisions in searching
through the prior art and in the consideration of
the precise wording of the claims. If a reference
is missed, and as the Commissioner says sometimes
it is, that is unfortunate and can lead to all kinds
of economic loss. So, it is important that the
search be complete.

When I first entered the patent field as a young
searcher in my father's office, and worked in the
search room of the old Patent Office there had been
issued only a little more than a million United
States patents and there were then considerably
fewer foreign patents than are now in the Patent
Office files. The books of reference and the maga-
zines which have been published since then would
probably fill all the rooms in the State Department
building. Technological publications are piling up
at an increasing rate and the disclosures become
more and more complicated. And, as the Com-
missioner said, the complications increase.

To continue a little longer in retrospect, while I
am not going to take much of your time, I will add
that on some of the searches which I made in im-
portant cases, I would spend a long time. I some-
times found that the Patent Office files were not
complete. I would search through all of the perti-
nent patents, examine all of the likely looking books
in the library, and would then be left out on the
limb, so to speak, not having been able to produce
the desired anticipatory matter. And on one oc-
casion I can recollect going to the Bureau of Stand-

ards library and discovering there in a textbook a
reference to a solution of a problem in gear cutting,
which was what I had been looking for all through
the Patent Office and did not find. And another
time I searched the entire Patent Office with no
fruitful result and then wandered over to the Bureau
of Public Roads and found the sought after infor-
mation in one of their little publications. Another
time I found that the Congressional Library was
the best place to search, going into their stacks by
special permission and pulling down books one
after the other. Now if we are going to have a
system in this and in other countries, of efficient
examination, some mechanism, some procedure,
some method of assembling the information which
is contained not only in patents but in publications,
must be devised. And that system, of course, must
be one which can be tapped. There is no use in
accumulating such technological 'information if you
cannot use it. So that is your problem, gentlemen.
I know that it is of great interest to each of you who
come here from abroad, because in my visits to Eur-
ope, I have made inquiry as to your interest, and
I know that it exists. I am sorry that by reason of
the fact that there is another conference now taking
place in Rome, some of the other patent offices are
not represented here, their representatives having
gone to that meeting. I have met with the greatest
cooperation in my travels from those employees of
other Patent Offices who were studying the problem
which we are to discuss today. I have tried to work
the British system in the British Patent Office, the
punch cards and so on, without much success I can
say. I know that the need to improve the search
system is well understood in London.

Gentlemen, I will close by wishing you well, and
by expressing the hope that this meeting will be a
most fruitful one. As the Commissioner said, this
is not a problem which the United States alone can
solve. We need the cooperation of all our friends.
We well know that in Europe and in other countries
of the Free World, there are many extraordinarily
well-versed, knowledgeable, scientific minds which
can be brought to bear on this problem so that, by
the exchange of information between those here in
the United States and the knowledgeable persons in
various other countries which are friendly, and in-
terested, we can solve the problem eventually. It
may take quite some time to accomplish our objec-
tives and I think we should not be engaged in this
effort under the impression that it can be solved in
any great hurry. But we have already made prog-

ress, and I am sure that some day we will have
the answer. Thank you very much for having
listened so patiently.

Chairman Fay: Mr. Watson, I want to thank you
and Mr. Ladd very much for this fine introduction
to our workshop.

I wish you would examine your programs, if you
have not already done so and see the subjects that
are coming up. You may wish to add to the pro-
gram after you have had a chance to study it and
see what other matters you would like to discuss.
Now I want you to know that this is your meeting.
This is your workshop, and we will be glad to take
up any order of business you wish.

Before we adjourn for a few moments, I would
be very pleased to entertain any suggestions you
might have as to topics you might like to add to the
agenda or things that you would like to discuss when
we return here at about 11:10. Are there any
suggestions? If there are not, we will adjourn until
11 : 10. I would suggest that you introduce your-
selves to each other and begin any discussion that
you would like to have with each other, and then
we will come back and see what comes of it.

A short recess was called by the Chairman.

Chairman Fay: We have some slack in our program
this morning. Therefore we did a little extemporiz-
ing and came up with this suggestion. We would
like to outline the problems, or some of the problems
facing the U.S. Patent Office. We have heard from
Commissioner Ladd and from former Commissioner
Watson as to the nature of the problem, and I would
now like to call on some representatives from our
office to set forth, in a small way, some particular
problems, so that we may understand some of this
challenge in a little greater depth. Then after
hearing from these people, it may be that some of
you from other countries or from business or from
universities would like to outline to us your problems.
We have hastily looked about and would like to in-
vite Mr. Irving Rotkin, a Patent Classifier in Divi-
sion 93 in the Patent Office, whose work involves
metal working and the mechanical arts, to give an
impromptu discussion of some of the problems.
Mr. Rotkin.

Mr. Rotkin: I have just scribbled seven lines of notes
so I can assure you it will not be too long. The
search problem that faces an examiner in the Patent
Office, as well as someone searching for an attorney
or for an inventor, is formulated by the claim. As it
is the claim that determines what an inventor gets

when he receives a patent, the claim is the yard-
stick and the measure used in weighing the patent
both in application form and as Letters Patent.
Thus, when a man is searching in patent files or
literature, it is a claim or claims, that he is trying
to find. The statutes in this country require novelty,
utility, and inventiveness. Utility however, in gen-
eral is not something one searches for, and it poses
no search problem as such. Novelty, which is the
primary thing we search for, has several aspects
that I think are of interest.

For one thing, entry into our classification system
is determined in part, by the way an item is set
forth. For example, take the ashtrays that are on
the tables in front of you. If that is what is named
in the claim, you have a hint. You are going to
look for a new ashtray. If one does not call it an
ashtray, but calls it a receptacle, you have a slightly
different problem as to where to look and how to go
about it. One may not call it a receptacle; it may
be called a paper weight. The fact that it looks
like an ashtray and can be used as an ashtray is a
hint to an alert searcher as to how and where he can
get into our system and start looking for similar
structures. He may look for it by name or by what
his senses tell him it may be used for. For example,
there are ashtrays and paper weights that, at least
according to detective literature, have made excel-
lent weapons (though I would not recommend
searching the weapon art for an ashtray ) . I merely
cite this to indicate that there are many ways of
getting into a related classification area for search-
ing.

In terms of novelty, the question is are you going
to find an exact copy — an exact duplicate? Are
you going to find differences? How far will you go
in accepting these differences? In the United
States patent system the statutes permit very broad
terminology. The device or item can be claimed
in terms of means; i.e. a broad means to accomplish
a particular function. This requires a look into the
concept of equivalents, and again it is a search
problem in the sense of how far afield one will go
for an equivalent. If a claim reads "means to
secure", what is a reasonable equivalent? Where
a nail has been shown, is a screw a reasonable
equivalent? When is a screw not a reasonable
equivalent? This will depend upon the utility as-
pect, which itself is not a search problem, as to
whether a screw is a proper equivalent for a nail
or whether a bolt is an equivalent. Is a spot
weldment an equivalent? Is soldering an equiva-

lent of welding, and so on. The question of how
far afield a searcher will go before saying that
something is no longer an equivalent determines the
scope, extent and/or time of search, assuming that
finances are not entered into the picture.

One also is troubled by the fact that what may
be an equivalent in one field of endeavor may not be
recognized as equivalent in another field of en-
deavor. At what time did a transition occur? In
the United States if a man claims a particular struc-
ture, and you find the exact copy of that structure
in every sense of the word, ( i.e. it is congruent, and
you can superimpose one on the other and they are
identical) it does not matter what it is used for.
If that is what he is claiming, you have completed
your search. But if he has any difference at all,
when will you worry about the difference? And
here we enter into another area which creates a
problem to the searcher in the United States patent
system, and that is: to receive a patent "invention"
must also be present. Invention as far as I am
concerned, and it is purely a personal approach, re-
flects a state of mind. It represents a difference
which one knowing the problems in a particular
area, recognizes as being sufficiently different to be
worth a 17-year exclusive grant by the United
States Government. People may differ with me as
to what invention is. I find this the handiest way
of defining it. A man technically qualified and
sufficiently versed in the legal lore that has grown
up around the term "invention" will look at the
difference and say this difference is great enough to
be worth 17 years of exclusive grant. But, you see,
this determines how far one is going to search. Are
you going to look at a difference and say, "Well,
yes, it is different, but it is not enough different to
be worth a 17-year patent. End of search!" Or
are you going to say, "I do not know whether it is
worth a 1 7-year patent ; I am going to search until
I find it." Every searcher must face the problem
of how short he is going to make his search or how
long.

A searcher is also faced, particularly in the chem-
ical area, with the question of variations. One may
claim "means for sitting" — e.g. a chair or stool. He
may then claim variations on that theme, and assign
a particular configuration to the chair with a very
broad claim reading on that configuration, and fur-
ther variations in detail of that configuration. Now,
as a searcher, is one satisfied in finding just the broad
description of the chair, or must one go into the
various species, the various variations in the chair?

It works both ways. One may find a particular
variant as set forth in an application and have to
decide whether this closes the search or whether one
must then continue in order to find all the variants,
as well as possibly a teaching of the generic or gen-
eral terminology to describe that entire family or
that entire class of chairs. In many instances the
classification systems is such that non-semantic
variants are all in the same pile, the same sub-class,
the same class. They may be so distributed in the
chair art, but in most instances as soon as the claims
vary one from the other the search varies to pretty
much the same extent. Unless one is to start with
Patent No. 1 and search through Patent No. 3,000,-
000, the searcher must determine by using the classi-
fication tools as best he can, where he will have the
optimum search experience.

Somewhat the same search problem, and one to
which I am much closer because it relates to the
mechanical field, is posed by claims that are couched
in combination and sub-combination relationships.
In searching for an automobile, this is fairly straight-
forward. There is a class of vehicles. But if one
claims, in addition to the automobile and its various
component parts, details of any particular part of
the automobile, there is literally no limit as to how
far afield the search can go. A transmission is not
in the same class with an automobile. Gears that
make up the transmission may be in an entirely dif-
ferent part of the class. And if one goes into further
detail in terms of how the gear is made or what the
gear is made of, one is still further afield from the
original class of automobiles.

What I am trying to indicate here is merely the
fact that a claim, being our initial starting point,
determines how far afield one must search to meet
the claim, and in meeting it, you have to meet it in
terms of whether it is new, or not new, because if it is
new, that is the end of your search. If it is not new,
is it sufficiently different so that you think it is
worthy of a 1 7-year grant? And again, if it is new,
is there something old that is an equivalent? If the
old teaching is such that reasonable people in re-
lated fields of technology will recognize and admit
that one is an equivalent for the other, even though
it is new, again a searcher can stop searching. But
I want to caution you that in the present state of our
classification, almost every time you talk in terms of
searching for a variant or a difference, a graduation
in the claim language, you are generally off at
another tangent in the 3 million patents that make

up our American patent system, as well as the 9
million foreign patents we now have.

The only thing I would like to add is truly taking
advantage of having you here before me. I am
hoping that in the discussions today, as well as
tomorrow, there will be a reasonable amount of
emphasis on the problems of today relative to the
manual classification system, as well as looking
forward to mechanisms or machines for aiding
searching tomorrow. Thank you.

Chairman Fay: Thank you very much, Mr. Rotkin.
Are there any questions? If there are not, I am going
to ask another of our examiners, who works in the
electronics area, Don Stein, to tell about some of
the problems he has encountered in his work in our
Office of Research and Development. Mr. Stein.

Mr. Stein: Thank you. I will try to tell you a little
bit about what the Office of Research and Develop-
ment in the United States Patent Office is doing to
help solve some of the problems in storing and
retrieving electronic patents.

We have many sub-projects within the electronics
project, reflecting a balance of shorter and longer
range objectives. Let me describe first our major
activity over the past year which is of a shorter
range nature. We have selected a group of patents
which include miscellaneous transistor circuits. We
have attempted to devise an indexing scheme that
would lend itself to mechanization for this group
of patents. We drew up a list — a schedule, a classi-
fication list — of approximately 1 ,000 terms to cover
the information in these patents. We are talking
now about approximately 1,500 United States pat-
ents, and approximately an equal number of foreign
patents that are classified in this particular area.

Each patent is analyzed by human beings and
is analyzed three different times: two analyses de
novo and one reconciliation which will brina: to-
gether the two analyses and is actually a third
analysis which reconciles the first two. Each anal-
ysis takes an average of approximately two hours
to complete, so that the total analysis time for each
patent is six man hours. Each patent is assigned
approximately 100 to 125 index terms which reflect
in great detail the information contained in the
patent.

At present we have completed analysis of approxi-
mately 700 original United States patents which are
in operation and actual use by examiners in search-
ing. The searches are performed manually by using
Batten or peek-a-boo cards. There is a great deal

645803 0-63-2

of flexibility and redundancy in the system, so that
the patents may be retrieved from many different
points of view. There is a burden upon the searcher
to vary his search request as he is searching. As he
sees what kind of answers he is getting, he is ex-
pected to broaden or narrow his search, or perhaps
change the point of view of his search, so that his
search will be complete. This briefly describes .his
aspect of the electronics work.

Now, using this file as a basis, we are beginning
to perform research in two areas that will attempt
to mechanize the analysis phase, which is the bur-
densome phase, as you probably recognize. We are
first going to study the natural language text of
these patents and the possibility of retrieving infor-
mation directly from the natural language text.
And secondly, we are studying the possibility of
encoding and retrieving the electronics circuitry
which appears in these patents.

For the experiment with the natural language
text, we are selecting 100 of these transistor circuit
patents. From the natural language text of these
patents, we will produce a magnetic tape which will
have full text of each of these hundred patents.
This magnetic tape will then go on a computer and
we will attempt to determine what techniques are
necessary to retrieve the same concepts, using a
computer and the natural language text, as are
encoded by manual means at present. In other
words, we will select certain concepts which are
coded manually, such as "feedback" or "gating"
and we will attempt to retrieve them from the nat-
ural language text. We will attempt to draw up
rules for retrieving them from this text. We wish to
determine how feasible this approach is for elec-
tronic patents.

With regard to the circuit itself, I would like to
note two things: First of all, we have performed
some preliminary experiments which seem to indi-
cate that a great deal of the circuit information is
contained in the text itself. We had a student sum-
mer employee who was given approximately ten
patents without the circuit drawings, and he was
asked to read the text and reconstruct as much of
the drawings as possible. In approximately seven
of these ten cases, he was able to reproduce the
entire drawing, except for one or two connections.
In the other three cases he was able to reproduce all
the connections which were essential to the inventive
concept, leaving out certain environmental connec-
tions, which were not completely described in the
text. This would seem to indicate, at least pre-

liminarily, that it may not be necessary to fully
encode the circuit itself, but it may be possible to
work with the textual or natural language material
in the patent.

Secondly, with regard to the circuits, we have
initially explored three different methods of repre-
senting the entire circuit by automatic means. The
first is a linear notation method. The second is a
matrix notation method. And the third is a lattice
method — a mathematical lattice type of represen-
tation. We are at present attempting to carry
these three schemes to a point where we can test
them to see to what extent it is possible to retrieve
portions of drawings automatically, using machines.

Perhaps at this point I should ask if there are any
questions as to any of the specific details of our
work.

Mr Cole: Do you look at the claims at all when
you. analyze this material, to get any help from the
claims as to what the novel matter is which will be
classified, or do you just go ahead and classify the
whole business?

Mr. Stein: I presume that you are talking about
the manual analysis. Inasmuch as the United States
Patent Office grants a patent on novelty, and a
patent application may be rejected not only on
what is previously claimed, but also on what is
previously disclosed, it is necessary to read and
analyze the entire patent text, not only the claims.
We do pay particular attention to the claims in the
manual analysis, and insure ourselves that every-
thing which is in the claims is properly encoded on
our system. But we also, in great detail, read the
entire specification and look at the drawings when
making this manual analysis.

Mr. Lawlor: Have you tried giving some exam-
iners the drawings and asking them to reproduce
the specifications?

Mr. Stein: No, that we have not tried.

Mr. Jancin: I'd like to ask a question directed to
Mr. Rotkin as well as yourself. As I see it, there
may be two problem areas here. One might be
looked upon as the long range objective, when
mechanized searching will be a very neat kind of
thing with little neon lights blinking and cards hope-
fully passing here and there, and then we have a
more immediate problem — the problem that exists
today. I would like to toss out this general question
to the Patent Office R&D, which I think is prob-

ably more knowledgeable in this area than any other
group. Have we taken a look, or have we made
any attempt to look at the existing hardware, with
the possible assistance of programmers in order to
come up with a short range — I won't call it solu-
tion — but short range assistance, to help ourselves
today with something that is not as sophisticated
as we might like, and then working on the side in a
more truly research effort, toward a more sophisti-
cated goal some years ahead? Could anyone answer
that please?

Mr. Rotkin: This is purely a personal answer. 1
can not pretend to speak for anyone else. Working
within the confines of the budget and personnel, the
effort that has been made has been very limited.
There is no denying the fact that the simple punched
card or the simple McBee sort system, could be
useful to a limited extent in many areas. Some
examiners in the Patent Office have seized upon
these to help them, once they have gained some
experience in their particular art. I personally feel
that this could go further in many areas. This
would simply be a manual card collection and man-
ual pin-sticking. In some areas, there is no reason
why it could not go beyond that, particularly as
the chemists have discovered in their area. This is
without any theoretical advance over what we know
today either as to technique or equipment. But as
I have said, because of a limitation of money and
manpower, and the press of other work, the effort
now has been to continue with our present system,
and work on what I call the "pie in the sky"
approach of mechanization tomorrow. I am hop-
ing that our Office will devote more attention to
the immediate problem, but this requires both man-
power and money.

Mr. Pfeffer: I would like to answer Mr. Jancin
from the point of view of the Research and Develop-
ment group. We have adopted the viewpoint of
looking both at the long range and the short range
problems, with a view to trying to solve some of the
more immediate problems with respect to helping
the examiners right now. In addition to trying to
solve some of the more difficult long range problems,
we have taken a good hard look at available hard-
ware through the assistance of many business or-
ganizations — your own, the International Business
Machines Corporation has been very helpful in the
past — and also through the National Bureau of
Standards, Data Processing Division.

The problem today is not a problem of hard-
ware. It is an intellectual problem. We have to
learn how to organize the material before we can
make use even of existing hardware. And the
hardware and the programmers by themselves will
give us no help whatsoever until we have learned
how to solve the intellectual end of the thing.

Mr. Stein: I would like to add that I think very little
is gained by mechanizing at present, or by utilizing
the present state of the art of mechanization. This
would be essentially mechanizing something similar
to the existing classification system, or mechanizing
a file such as we have done in transistors. Let me
add, the main thing we have gained in our tran-
sistor file is in the accuracy of search and in depth
of analysis. We have essentially encoded several
things which are not obvious by looking at the
patent in a cursory fashion, but are buried deep
within the patent. But I think on the whole very
little is to be gained by this.

I have just returned from a conference spon-
sored by the Association for Computing Machinery
on Techniques of Programming Digital Computers
for Information Retrieval. I think that conference
pointed out very clearly, that the main problem
areas today do not lie in devising specific programs
for computers, do not lie in the hardware area, do
not lie in applying the present state of the art to
patient office searching problems. Rather, the es-
sence of the problem — I think this was the consensus
of the people at this conference — is the intellectual
one. What is information? How should informa-
tion be organized? In what fashion can the
thoughts of the person who wrote the words be made
available to those people who are searching? I
think it is most important to solve this problem,
and that mechanization along the way will probably
not gain too much speed, and perhaps will gain
some accuracy. But I doubt whether it is worth-
while on a large scale.

Mr. Bulder: My question concerns three methods
of analyzing drawings, first the linear and second
the matrix. What do you mean by that? What
is a linear analysis? What is a matrix analysis,
etc.?

Mr. Stein: I can at present explain this only brieflv.
Let me, before I begin to do this, extend an invita-
tion to any of you to visit our Office of Research
and Development. We will be very happy to go
into as much detail as you might desire concerning
our present work in any of the projects.

Very briefly, a linear notation attempts to set
down a set of symbols in a linear fashion which will
represent the circuit. The matrix approach at-
tempts to set up a matrix of symbols which will
represent the circuit. And the lattice approach
attempts to set up an information lattice with
various information points, which would represent
the circuit. I feel I can not go into more detail
than that here.

The main problem here is that you have essen-
tially a three-dimensional array in an electronic cir-
cuit, and within the present state of the computer
technology, it is impractical to search every part of
every circuit in the file that is stored, in order to find
an answer for a question. One must devise some
shorthand technique of representing the circuits.

Mr. Wessell: Did I understand you correctly to say
that you used 125 terms for each patent?

Mr. Stein: That is the average, yes.

Mr. Wessell: Are these terms which appear in the
patent or do you apply some of these terms from a
pre-formed dictionary?

Mr. Stein: This dictionary was drawn up after a
careful study of selected patents from this file. And
the dictionary is constantly being added to and
modified as further patents are analyzed.

Mr. Wessell: And do you find this a very deep type
of index?

Mr. Stein: We feel that it requires individuals very
well trained in electronics, and also trained in the
United States standards of patentability, in order
to perform these analyses. We cannot use clerical
people, and we cannot even use electrical engineers
unless they have some training in the United States
standards of patentability.

Mr. Wessell: In searching then, you would have to
describe your question in very great detail, would
you not?

Mr. Stein: No. The terms are of varying scope.
Some terms are very broad. Other terms are quite
specific. So it is possible for the searcher by selecting
not more than perhaps ten to fifteen terms, and
modifying these as he is making a search, to narrow
his search down to a practical limit.

Mr. Michel: While manually extracting the data or
the material disclosed in the specification, does the
operator take account in every patent which he trans-
fers to a card, of the material which is obviously old,
which he knows has been encoded before in con-
nection with some other patent?

Mr. Stein: Yes. He encodes all the information be-
cause, although the individual material, or a specific
part of the material, may be old, that material in
combination with some other features of the inven-
tion may be new. So that we encode everything that
we can find there.

Mr. Pfeffer: May I answer a little bit further? Com-
missioner Michel, in addition to the statement which
Mr. Stein gave you, it is very difficult for any indi-
vidual to determine all by himself what is old and
what is new. And until we can find some way of
correlating information and making a pass through
it to make determinations of this type, it is safest at
this point to encode everything that is present. We
have had many experiences where the examiner
thinks that something is old. It looks old. He may
have seen it somewhere and it may have been in a
pending application rather than a published
article. Secondly, there is always the possibility of
losing information which may appear somewhere
in some obscure journal or publication and nowhere
else. So it would be helpful to have that informa-
tion made available again, especially where the
publication date may be critical.

Mr. Michel: Even though that might produce a
whole lot of duplication in your cards?

Mr. Pfeffer: Yes.

Chairman Fay: If there are no further questions, we
will recess until after lunch. The afternoon session
will begin at 1 : 30. Thank you Mr. Stein.

I want to make a couple of announcements. I
have been asked to advise you that at luncheon
today there will be no head table and no speakers.
Our visitors should feel free to take any of our men
who are willing to talk and ask them any questions
you feel you want to. And I will ask our examiners
to disperse themselves among our visitors so that we
can get the full benefit of this interchange.

The opening session was adjourned.

Second Session

Monday Afternoon, October 23, 1961

The second session was called to order at 2
o'clock by Mr. John Green, Director, Of-
fice of Technical Services, Business and
Defense Services Administration, U.S. De-
partment of Commerce, acting as Mod-
erator

Mr. Green: I am with the Department of Commerce
and I have been asked to moderate this afternoon's
session. As you know, this is the first technical meet-
ing of this group and we are endeavoring to identify
areas of cooperation in which the various patent
office representatives might work together to ad-
vance the cause of storage and retrieval on patent
information. Of course, in today's session we are
speaking only on patent classification, and we have
three authorities who are going to make direct pres-
entations. As you can see from your programs, they
are the Honorable Werner Rubach, Director of the
Technical Section of the German Patent Office;
Mr. H. S. Vincent of our own Patent Office — The
United States; and the Honorable L. F. W. Knight,
Superintending Examiner in the United Kingdom
Patent Office.

Our plan this afternoon is to have these presenta-
tions without a break for question and answer.
Then we will have a roundtable discussion, which
will last, I imagine, until about five this afternoon,
at which time we will conclude. We hope to have
a short intermission for coffee around four, and
there will be proceedings of this meeting printed
for those of you who want to keep abreast. Every-
thing is being recorded.

With that short background, let me proceed to
our first speaker, who is Mr. Werner Rubach. He
is Director of the Technical Section in the German
Patent Office. As such, he has responsibility for
examination, classification, documentation and the
assignment of applications. His educational exper-
ience includes a mechanical engineering degree
from the Technical University of Berlin, and he
will speak to us this afternoon on the International
Patent Classification System. Mr. Rubach.

Mr. Rubach: Ladies and Gentlemen, I will speak
to you about the international classification of pat-

ents for inventions. Systems of classifications are
tools to help us find our way through an overwhelm-
ing variety of subject matter. If, as in the fields
covered by the Patent Offices, new varieties of sub-
ject matter are coming up incessantly, constant
effort will be spent in order to keep our tools keen
and efficient.

In Europe such an effort is being carried out in
parallel at present under the European Convention
on the International Classification of Patents for
Invention, the Preamble of which reads: "The Gov-
ernments signatory hereto being members of the
Council of Europe, considering that the aim of the
Council of Europe is to achieve a greater unity be-
tween its members, for the purpose of facilitating
their economic and social progress by agreements
and common action, and having regard to the reso-
lution of the Committee of Ministers of the Council
of Europe of 12th September 1952 recognizing the
necessity of the generalization of the practice of
examination of patent applications for novelty, and
considering that the adoption of a uniform system
of classification of patents for invention is in the
common interest and likely to contribute to the har-
monization of national legislation have agreed that
each contracting party should adopt the system of
classification of patents for invention set out in the
Annex to this Convention." The abovementioned
uniform system is being used at present by the Pat-
ent Offices throughout Europe inasmuch as their
Governments are members of the Council of Eu-
rope. Both Belgium and France apply it as their
exclusive means of classification, whereas all other
countries concerned use it as secondary classifica-
tion aside from their national ones.

The outline of the uniform system will be found in
the Annex to the Convention where eight main clas-
sification sections are listed, namely, (a) human

necessities, (b) performing operations, (c) chem-
istry and metallurgy, (d) textiles and paper, (e)
fixed constructions, (f) mechanics, lighting and
heating, (g) physics, (h) electricity. For further
subdivision of the fields covered by these sections,
about 103 classes are provided, of which 52 classes
have already been elaborated by a Working Party
of Experts on classification. Let us stress here that
the sub-classes of the I.P.C. embrace a far broader
field of technology than sub-classes in the U.S.
classification system. To cite an example, while the
international class F-05 deals with all pumps, the
various known types of pumps are grouped in sub-
classes of that international class, each sub-class
again being split into many groups and sub-groups.

The Working Party of Experts on
Classification

Under Article Two of the Convention the re-
sponsibility for further elaborating the 103 classes
of the Annex, is assigned to the Committee of Experts
on Patents of the Council of Europe. Other respon-
sibilities of that Committee include the development
of proposals for the unification of the patent laws
and procedures for granting patents in the various
member countries of the Council of Europe. For
work on the International Patent Classification, the
Committee of Experts on Patents have legalized the
Working Party of Experts on Classification whose
activities began back in 1952. This Working Party
includes experts of France, The Netherlands, the
United Kingdom, and from the Federal Republic
of Germany, the Chair being held by the repre-
sentative of a country not concerned with the elabo-
ration of the scheme at present. A member of the
Swedish Patent Office acts as Chairman. When in
1952 the Working Party began their discussions at
Strasbourg, they at first tried to create an entirely
new system for subdividing the whole field of tech-
nology. A draft listing 115 classes was set up and
disseminated to the member countries of the Council
of Europe, for consideration. However, no country,
or no patent office respectively, could agree with the
system proposed. On that, it was decided that a new
international classification system should rather be
shaped in close correlation with one of the existing
systems of patent classification, in other words,
with a system already in general use all over Europe.
This condition was met best by the German System
of classification, then in use without any alteration
in the Scandinavian countries, with some alteration
in the Netherlands, Austria and even in Russia,

Poland, Czechoslovakia and Hungary. Even now
it is used in Switzerland. It was, therefore, even-
tually chosen the system underlying the scheme of
the new international classification.

When the activities of the Working Party started,
after the previously mentioned difficulty, the broad
subdivision cited above was worked out for good
in order to cover the whole field of technology. It
is used today throughout Europe, so that you will
find a corresponding symbol on every European
specification issued by a member nation of the
Council of Europe.

Principles Guiding the Working Party

Classification for invention may be conceived
under several aspects, namely, subdividing the sub-
ject matter according to construction, function or
purpose on the one hand, or to an industrial branch
or art on the other hand. I do not know of any
workable system in any Patent Office which would
be built up solely on either of the two. They rather
incorporate both aspects at the same time. It is
true, however, that some countries seem to prefer
the construction, function or purpose, while others
favor systems reflecting subdivision of industry in
branches, or of technological arts.

The difference is well illustrated by a comparison
of the classification systems presently used in the
United States and the German Patent Offices.
Construction, function or purpose prevail by far as
characteristics of the main units of subdivision on
the American side, i.e. the classes include subject
matter related by technological functions, whereas
the German practice subdivides according to art.
There are, however, some classes also in the German
system where the subject matter is kept together
according to fundamentals either of construction,
function or purpose. If the German main classifi-
cation unit is set up that way, it is called a general
class. There was no doubt that a certain percent-
age of the German classification system could not
be taken over into the new European scheme with-
out checking, amending, and reshaping according to
the latest developments of the arts. Taking into
account such modern trends as standardization and
automatization, it was felt to be necessary to have
more general classes in the new European scheme
than formerly provided for in the German system
for patent classification.

As a consequence, classes devoted to a special art,
which I might call special classes, are now being
restricted to subject matter really and truly designed

for a special art, or adapted to a special use. All
such elements or construction units, however, which
perform a distinct technological function in a
machine, in an apparatus, or in a device and which
are not adapted in their constructional features and
do not differ from the common constructional
features of such elements, should be brought to a
general class. In cases of doubt, the subject
matter will be classified under the relevant general
class, or in other words, the general classes take
precedence over the special classes.

Procedure

If another one of the classes listed in the Annex
is scheduled to be subdivided, work thereon begins
with a first proposal to be submitted to the Working
Party by one of the countries practicing novelty
searches when examining patent applications. The
first proposal which, of course, takes note of any
current development of the art concerned, will be
disseminated to the members of the Working Party.
They study it and are free to suggest alterations or
amendments and to make other remarks to the
author of the first proposal. All suggestions are
now scrutinized by the author and amendments
and changes felt to be useful are incorporated into
a revised draft which, again, is sent to the members
of the Working Party. In due time, the revised
draft will then form the basis of discussions in the
forthcoming meeting of the Working Party. Such
meetings are held twice a year, Spring and Autumn,
with a view to finalizing and legalizing the revised
draft.

Progress

As there are various principles applicable to
classification schemes, as in practice systems exist-
ing in Europe show a lot of differences, and as
furthermore, opinion might differ as to how to
classify patents on inventions granted under differ-
ent laws and procedures, one can imagine the
extreme difficulty faced by the members of the
Working Party in their current work. Under these
circumstances, the subdivision of classes has not yet
been terminated. Moreover, I do not hesitate to
admit that progress in the sessions of the Working
Party is now slower than it was in the beginning.

At that time we could go ahead without much
regard for subject matter dealt with in already
elaborated classes. At the present period, however,
when more than half of the subdividing work has
been completed, much care must be taken to avoid

overlapping of groups and subgroups, as the finer
subdivisions are called, with groups and subgroups
in other classes. It means more and more time
spent on exact wording and reliable cross refer-
encing. Much discussion has been devoted to ways
and means of accelerating the work. But so far
no sensible solution of the problem has been found.

Rules Governing Classification of Products or
Processes, Special Purpose Machinery
and Special Purpose Methods

In dealing with the problem of where to classify
products or processes for which no classification unit
can be spotted that would be immediately appropri-
ate, the following interpretation of the International
Patent Classification scheme has been agreed upon
by the Working Party. To read the agreement, "In
interpreting this system of classification, it should be
noted that the wording describing the subject matter
of a classification unit normally designates either
a product or a process, or both, the words "product"
and "process" being used here in their widest pos-
sible sense, as indicated by the examples given below.
Wording in other terms, e.g. designating a use or
application of a process or material, is used in some
cases.

"Examples of products are : chemical compounds,
compositions, wire films, fabrics, articles, apparatus,
tools, machines, transport means, structures, assem-
blies, electric circuits and systems composed of these.
Examples of processes are: polymerization, fermen-
tation, separation, shaping, conveying, treatment of
textiles, transfer and transformation of energy, build-
ing, preparation of foodstuffs, testing, methods and
modes of operation of machines, and processing and
transmitting information. In many cases, the sub-
ject matter of a classification unit is not confined to
the product or process designated, but should be
understood as follows: ( 1 ) A classification unit des-
ignating a product includes the process used in the
manufacture of the product unless these processes
are covered elsewhere in the system by a classification
unit or units designating a process. Insofar as the
unit thus includes a process, paragraph two (which
follows) also applies to it. A classification unit des-
ignating a product also includes any parts of the
product not covered by other such units. (2) A
classification unit designating a process includes any
apparatus tools, machines, devices or materials used
in the process, unless the apparatus, tools, machines,
devices or materials are covered in the system by
units designating products; and also includes any

product produced by the process, unless it is covered
in the svstem bv a unit designating a product". This
interpretation has proven to be quite efficient when-
ever new products or processes have been involved.

But how do we handle the cases where machinery
or processes are adapted to special purposes? It is
our practice to classify the use for a special purpose
of a machine, an apparatus or a device under the
class or sub-class provided for this purpose. In the
case of substances or materials, however, a tricky
problem remains to be settled by our final decision.
Where do we want to put an invention dealing with
the method for making a substance in such a way
that a substance with definite special physical prop-
erties results? Such cases could, indeed, be asisgned
either to the general class provided for such sub-
stances, or else to the class grouping all objects for
the function of which such properties are used, only
just these properties guaranteeing that the desired
function really occurs.

On this question it has proven to be extremely
difficult to come to an agreement. Personally, I am
inclined to favor a solution that would gather and
keep together under one heading all methods for
making substances or materials, even the particular
methods that include steps for obtaining products
with definite, special physical properties.

Application of the International Patent
Classification

The Annex to the European Convention, where
the basic systems of classes and subclasses is set forth,
in its preliminary note explains both underlying
principles and application as follows: "This classi-
fication scheme is devised only for the purpose of
classifying inventions which are the subjects of
patents, and the requirements of Article 3(1) and
3(3) according to which the printed patent speci-
fications issued by the Contracting Parties have to
be marked by the national authorities with the com-
plete symbols of the International Classification, are
intended to be applied only to such symbols of
classification as are necessary to classify the invention
of any particular patent. These symbols will usually
relate only to one subclass and one place in the
elaborations thereof, but in certain cases they may
extend to two or more subclasses and/or places.

There is, however, no bar to the use of additional
symbols of the classification for any other purposes
(for example, for classifying subject matter which
may be disclosed in a patent specification without
forming part of the invention thereof), provided

that, if such additional symbols are placed on the
documents referred to in Article 3(1), these sym-
bols are distinguished from those which relate to
the invention as stated in the paragraph above, for
example, by being printed in lighter type. It is not
necessary for the additional symbols to include any
symbols relating to groups or subgroups of the
classification system (which are not yet included in
the" Annex to the Convention), and they may, if
desired in any case, indicate only the class
concerned."

There might be argument as to what purpose a
patent classification system is meant to serve. Even
in the Working Party, lengthy discussions have dealt
with the problem whether the International Patent
Classification is a tool designed exclusively for pat-
ent office practice, or whether it is meant to be
equally important for industrial and other circles
and enterprises interested in publications in the pat-
ent field. As a matter of fact, the latter users do not
only wish for an instrument with which to carry
through novelty searches, but they want to have
available also a tool for finding out where the road
is free for their own research and development, and
which roads, on the other hand, are closed by valid
patents belonging to others.

What does the Convention title say to that? It
reads : "European Convention on the International
Classification of Patents for Invention". So the
problem boils down to our deciding what a patent
really is. Is it a sheet of paper only, technological
literature to be called "specification", or is it some-
thing more than that, namely, a monopoly granted
on an invention found to be novel?

If the latter alternative given is held to be cor-
rect, every scheme for the classification of patents
represents a tool for the orderly grouping of rights
protecting inventions officially screened for novelty.
This is what we want to keep in mind when reading
the aforesaid paragraph ( 1 ) of the preliminary
note.

When classifying inventions along the lines set
forth in paragraph ( 1 ) of the preliminary note cited
above, the International Patent Classification lends
itself easily to use for multiple classification. For
this application several symbols will appear in print
on the specifications instead of only one, namely, as
many as there are classes or subclasses respectively,
relevant to, or covered by, the invention concerned.
Also, whenever subject matter characterized by a
subsequent claim relates to another class or subclass,

this multiple classification of the invention will be
used.

In German Patent Office practice, however,
every application is meant to be handled under the
exclusive responsibility of one examiner only. This
is the same person who eventually will be entitled to
sign the papers granting a patent. Under this pro-
cedure, one symbol on a specification stands for this
one and only person responsible in that case, and a
subsequent claim for which another symbol would
be appropriate will be taken out of the application
and examined in a separate application under its
appropriate symbol.

Aside from the symbols relating to the invention
proper, specifications may be indexed by using still
other additional symbols. Such symbols would re-
late to technical features mentioned in the descrip-
tion of the invention, but not forming part of the
latter. By this practice, the International Patent
Classification is, in fact, used directly for purposes
of documentation. And, indeed, any searching of
specifications which bear such additional indexes,
might be carried through faster than before because
the searcher, by simply scanning all classification
symbols printed on the head of the first page, al-
ready gathers conclusive information about the
scope of technical subject matter to be found in that
description.

For the time being, experiences in the fields where
the International Patent Classification has been
elaborated are rather encouraging, so that it seems
reasonable to expect eventual success of our work
conducted under the motto: "E pluribus unum."
Thank you.

Mr. Green: Thank you, Mr. Rubach. As I said
earlier, in order to move the program ahead and to
preserve adequate time for the roundtable discus-
tion, I hope you will all save your questions for
Mr. Rubach and his colleagues until later.

Our next speaker, Mr. Herbert S. Vincent, who
is a patent classifier in the United States Patent
Office, took his degree in Physical Sciences from
Ohio State. He has been in the Patent Office since
1942 and has specialized in the classification of
patents since 1946. He has undertaken the re-
classification of several of the classes in such diverse
fields as coating apparatus, laminated fabrics, and
electrical connectors. This afternoon he will de-
scribe to us the classification system employed by the
United States Patent Office. Mr. Vincent.

Mr. Vincent: It has been estimated that about half

the time spent in the examination of patent applica-
tions in the U.S. Patent Office is spent in the search
for references, the prior anticipating publications.
As the sciences grow more complex and our fields of
knowledge expand, the search problems increase, of
course. In our patent system alone we start today
with a mass of three million original patents, and
three million two hundred and fifty thousand cross
references. Toss 48,000 newly issued patents into
the pot every year. For each newly issued patent
add an average of two cross reference copies. To
this add the vast numbers of foreign patents, tech-
nical literature, publications, and miscellaneous un-
official references, and we find that we have a rapidly
growing child, one that might one day be big enough
and mean enough to lick the old man if he is not
kept well in hand and carefully brought up.

Today in our patent classification system we have
three hundred and eight main classes of invention,
broken down into 59,300 subclasses. The new pat-
ents which issue every week, are placed in the system
by the examiners who issue them. They are placed
in accordance with class and subclass definitions, and
by following definite rules of procedure. The classes
are developed by our Classification Division, whose
primary function is the creation of these new
classifications.

The purpose of a classification must, of course,
be the guide in its derivation. A patent classification
must primarily divide the subject matter into easily
searchable groups for patent examining purposes,
must reduce the danger of issuing interfering appli-
cations for patents, and must provide fields of search
for industry to ( 1 ) avoid infringement of patents
and ( 2 ) provide for state-of-the-art searches.

What is the best basis for bringing together into
major divisions, or major classes, those things that
are the nearest alike for our purposes? It has been
determined that utility, or use, must be the founda-
tion on which to build our major groupings. But
the utility must be the immediate utility, not an
ulterior, remote, or accidental one. The term
"proximate function", is frequently used by us to
describe this particular characteristic. A pipe with
fins on the periphery might be used as a condenser
for alcoholic vapors, or it might be used as a radiator
to heat a room. These two different uses for the
structure are accidental functions. The direct, or
proximate function, of such a finned tube is heat
exchange. A proper and worthwhile classification
for our purposes should be based on this, rather than
on the accidental functions. Thus, if we have two

patents, one of which discloses the structure for use
as a radiator, and the other discloses its use as a
condenser, the patents would probably be classified
together on the basis that they both have the same
proximate function, even though they have different
ultimate uses.

So we believe that our patents are, in the main,
better classified in accordance with the function or
effect they must perform or accomplish, rather than
in accordance with an object toward which the act
is directed or in accordance with the result which is
intended to be achieved. Of course, a word of
caution must be given here. A hammer, a golf
club and a drumstick all have the function of striking
a blow. But obviously, they should not be classified
together. So the test of special use might have to be
applied to the selected functional basis. Of course,
the prime function of the classifier is to make the
proper determinations as to the depth he will analyze
to determine the similarities in the proximate func-
tion of the patents that he is classifying.

Our classification then, at least in its larger
groupings, is not based on structure alone. For
minor subdivisions the classifier might well use a
structural basis of distinction. Having collected all
heat exchangers into one class, the classifier might
then subdivide them on the different structures that
are recited.

Our time today is limited so we can not go too
deeply into the theoretical basis for our classifica-
tion. Obviously, other factors must be considered.
In the manufacturing arts, for example, a plural
step process must be considered from the point of
view of producing a special effect or product. A
composition of matter or stock material should be
classified in accordance with its composition where
possible, although the test of special use might have
to be applied here, too.

Every patent has subject matter that is recited in
the claims and subject matter that is disclosed but
not claimed. It has been deemed advisable to
classify on the basis of the claimed subject matter.
The claims set forth that which is novel and dis-
tinguishes a particular patent from all other patents.
Thus, in basing the classification on that portion of
the disclosure which is claimed, we may best be
assured that the significant subject matter is being
considered. Of course, the unclaimed disclosure
may be as valuable as the claimed disclosure for
search purposes. Therefore, the classification
should provide for both. A patent is placed as an
original copy on the basis of the claims and a search

for unclaimed disclosure is provided by way of cross
reference copies.

Each of the 308 main classes are carefully and
meticulously defined and limited in scope as to sub-
ject matter. Further, each class is carefully
annoted with descriptions of related fields of search.
The classes are divided into subclasses each with its
definations and notes. The number of subclasses
within a class varies greatly, being dependent on the
nature of the subject matter activity in the specific
field, age of the class and other factors.

Among the 308 main classes, two guide lines are
used to determine placement of a patent into the
system. The first guide is a general superiority rule,
and the second is the specific class definition. The
general rule of superiority among main classes is
as follows:

Superiority of Types of Invention
( Claimed )

1. Process (of using a product)

2. Product (of manufacturing process)

3. Process (of making a product)

4. Apparatus (to perform (3) and/or to

make (2); e.g. machine,
tool)

5. Material (used in (3) to make (2) ; e.g.

composition, stock, blank or
intermediate)

The above listed items are arranged in order of
superiority, each item being dominant over those
following in the list. The list then is a hierarchial
determination for the placement of the original
patent.

This general rule of superiority, as set forth in
this list, is followed unless the specific class defini-
tions detail exceptions to the broad principles. If
so, the specific exceptions control classification.

So much for the distinctions among main classes.
Let us look now at a typical class and examine its
structure. I will now directly contradict myself
and say there is no such thing as a typical class. On
consideration I find that I have to agree with myself
that this is so! However, it has been found over
the years that in many new classifications a pattern
may be discerned, and this general arrangement
(subject to special exceptions which may be tailored
to any particular art) is usually followed in most
of our new classifications. In Figure 1, the letter
"A" represents the basic subject matter of the class.
By basic subject matter we mean the heart wood of

the class, the tie that runs through all the disclosures
in a particular class, the thing that relates all the
patents in the group. In Class 100, Presses, the
basic subject matter is the particular means for ap-
plying pressure to the workpiece. In Class 339,
Electric Connectors, it is the particular electrically
conducting joint that forms the basic subject matter.
In class 118, Coating Apparatus, it is the means
transferring the fluent layer of material onto the
base.

TYPICAL ARRANGEMENT OF A CLASS

(Basic Subject Matter of Class is Represented By the
letter "A")

I Special Types of A

II A combined with specified non-A features or

devices (A&B)

III A with perfectins features (A&P)
IV Convertible (A to or from B)
V Combined (A&X)

(no combined A below this line)
VI Plural A (A 1 with A 2 )
VII Types of A (A 1 — A")

VIII Subcombinations of A specialized to use with A
IX Subcombinations of A not specialized
X Miscellaneous

Figure 1

The titles I through X set forth in broad terms
the types of subclasses and also the general order or
scheme of superiority in a typical class. In this
arrangement, special or unique variants of the basic
subject matter appear in the plan as I. These are
considered to have some special quality of im-
portance which requires a high order of priority
and entitles them to appear in a high position in
the schedule.

Following will be II, subclasses relating to com-
bined features. These are devices of the class com-
bined with something that is not of the class, some-
thing in addition to the basic subject matter of the
class. Ill represents the basic subject with per-
fecting features, that is, something improving the
operation of the basic device or embellishing its
basic function. The distinction between II and III
grows thin at times and in fact it is not too im-
portant that a specific difference be drawn between

them. It is well, however, for the classifier to recog-
nize that the two types exist.

Many classes provide a subclass labelled "con-
vertible" as shown at IV. These are particularly
adapted to be altered to a different type of device.
An example of this is the kitchen stool that may be
converted to a ladder.

A miscellaneous subclass titled "combined" is
shown at V. This is the residual subclass for "com-
bined" features not provided for above this position
in the schedule. It should not be necessary to
search below this point in the schedule for com-
bined features.

Below this position, we deal with variants of the
basic subject matter of the class. VI sets forth
organizations involving multiples of the basic sub-
ject matter, followed by VII, the variant species of
the basic class subject matter. Generally the bulk
of the subclasses in a class are variants of VII.
Under VIII and IX are found subcombinations or
elements of the entire device and finally a subclass
titled "miscellaneous" (X) which takes anything
not provided for in any of the other subclasses in
the class.

So, in examining the stylized class you can see
that it has almost the structure of an inverted
pyramid. The special things, the complicated
things, the complex things, come first in our schedule.
We gradually work our way down through the sub-
ject matter, winding up with the less important
features, the sub-combinations, the things which
might be not of specific use, but of general use.

Figure 2 analyzes class 100, Presses in terms of
the abstract representations of Fig. 1 . In the column
labeled "subclass" are set out certain subclasses by
number and title as they appear in the class. The
right column gives the analysis of the subclasses in
the generic terms of the Fig. 1 representation. We
see that this class schedule is more than just a listing
of items. Just as a road map is more than a list of
places to go to. It is actually a physical plan with a
special hierarchical arrangement. The schedule
identifies relationships of concepts and things. It
sets forth the whole in position superior to the part
and the special effect before the general effect. By
setting up comparative relationships, it allows the
subject to be approached by comparison of similar
and dissimilar things whether or not a specific name
for the object to be sought is known.

Figure 3 is a portion of the Class 100 schedule.
Note the first line subclasses — subclasses 1, 35, 43,
53, 54, etc. These are capitalized to facilitate scan-

CLASS AND SUBCLASS SCHEDULE
ORGANIZATION TO LIMIT SEARCH

Subclass

Explanation

Special type of press
(A 1 ) as defined
above.

Superior to apparatus.

Automatic or

Any type press (A)

Material.

(except special type
binding taken out
above) with perfecting
features (Ap).

Interrelated or
Safety Controls.

A P .

Portable Receptacle

Special type of press

Lid Applying.

(A 2 ).

Compacting Ensilage

Special type of press

Within Silo.

(A 3 ).

With Additional

Any press A (except

Treatment of

A 1 , A' 2 , A 3 taken

Material.

out above) combined
with specified non-
press feature.

With Alarm, Signal,
Indicator or Test
Means.

A P .

100

With Ground Tra-
versing Wheels or
Guides.

A P .

101

With Cloth About
Cake Enfolder.

A P .

102

Residual subclass for
any combination
not specified above.

103
104

With Drain Means

Type of press A 4

for Expressed

(expressing type).

Liquid.

137

Plural Diverse Presses.

Plural diverse A.

144

Concurrent Pressing

Type of press A 5

and Conveying.

(screw press).

193

Plural Presses

Type of press A b
(sequential A 6 -A 6 ).

210

Roll and Platen . . .

Type of press A 7 .

211

With Flexible or
Deformable Pres-
sure Surface.

Type of press A 8 .

213

Interfiling Cups.

Type of press A fl .

214

Reciprocating Press
Construction.

Type of press A lu .

295

Platens or Pressure

Subcombinations

Surfaces.

peculiar to presses.

298

Cloth From Cake
Strippers.

Specified miscellaneous

299

Miscellaneous

igure

ning. They all have the same degree of indentation.
They are what we call coordinate subclasses. Also,
of course, subclasses 36 through 46 bear a coordinate

relationship to each other, although they are all
indented under subclass 35.

The claims or the claimed disclosure control the
selection of one of the 308 major classes for the
placement of the patent. Having found the class
which by definition can provide for the subject
matter, we can then proceed to place the patent in
the class schedule. One copy, the original copy
as we call it, being placed in the proper subclass.
Cross reference copies can then be added. The
classifier starts at the top of the class schedule and
scans down the coordinate first line subclasses, re-
jecting each until the subclass as titled and defined
includes the subject matter of the patent claim.
Having found the first line subclass, he then proceeds
to the coordinate subclasses indented thereunder, and
once more scans down, rejecting coordinate concepts
until the proper one for the claimed subject matter
is once more found. The process is repeated until
the ultimate coordinate subclass is found. Then the
patent may be placed into a subclass indented there-
under on the basis of disclosure (in this last step
only) if no indented subclass provides for the claim.
The reasons for relying on disclosure in this ultimate
step are twofold. First, anyone searching the genus
must search all of the indented subclasses anyhow.
So the patent will not be missed. And secondly,
anyone searching the indented species only will find
the original copy and thus, only one copy of the
patent need be placed. We have thus saved cross
reference copies which would overload our system.

Let us see how this works with a specific docu-
ment to be classified. Fig. 4 represents a patent.
This is a three-roll press. The material moves from
left to right, passes between two hips as you can
see, generated by the three roll. As it goes through,
it is squeezed once and then it is squeezed a second
time. Also disclosed in the document is the sensing
means, which senses the thickness of the material
coming out of the press and feeds back into the
drive. By controlling the speed of the drive you
can control the ultimate thickness of the material.
Between the two lower rolls there is a turnplate for
forwarding the material, helping it go through the
press. The top roll, as you can see, is spring-biased
downwardly toward the two other rolls and also has
a variable axis as shown by the arcuate arrow, so
that it can apply more or less pressure against either
of the selected two rolls. So far as the claim is con-
cerned, let us assume that what is recited in the left
column is what is claimed, and what is recited in
the right column is disclosed but not claimed.

Original Classification C. G. Smith

T. L. Stam, H. S. Vincent
Definitions in Bulletin No. 407

1952

1 BINDING

2 Methods

3 (Compacting and binding

4 With automatic or material triggered
control

With material winding or folding
With material severing
With material depositing or discharging
Binder applying

Preformed continuous annular binder
With precutting of binder to length
Plural flexible section binder
Through annular material
Helically along material
Indexing material between successive

placements
By rotation of material
By swing of material gripping jaw
Binder supported across passageway
for material
With material conveyer

With threader crossing material
path
Delivering binder to end gripper
Concurrently delivering to
joiner
Indexing gripper
Shifting gripper delivers

binder to joiner
Threaders
Curved guide channel for pushed binder

Material receiving loop channel
Orbit travelling binder placer

Reversing for successive placements
Binder tighteners and joiners

I Sleeve or clamp joining
Twisters
Binder tighteners
Joiners

Binder retaining material holders
METHODS

With pre-enclosing in textile

With separation from material of liquid

expressed
With heating or cooling
With cutting or comminuting
With winding or folding
Forcing through constricted passage
Successive compressions from different
directions
AUTOMATIC OR MATERIAL TRIGGERED
CONTROL
Of insertion of separator between

successive material charges
Of material addition, depositing or dis-
charging
Of parallelism of platens
Of roll separation or speed
Of actuating means

Material presence or position starting
By pressure applied to material
With dwell time control
Reversing
INTERRELATED OR SAFETY CONTROLS
PORTABLE RECEPTACLE LID APPLYING
With receptacle transfer means
With lid fastening
With gathering of contents
With lid locating means
Receptacle mountable
Upwardly displaced receptacle

We will make the assumption that this document
is properly classified in Class 100, Presses. We
now must locate the proper subclass in the schedule
for this particular document as an original patent
and also the proper place to cross-reference any of
the additional disclosures that we think worth classi-
fying. Following the rules, we analyze the claim,
in order to keep the subject matter firmly in mind.
We then proceed down the schedule scanning the
coordinate, first line, capitalized subclasses, reject-
ing each until the first capitalized subclass is reached
that can provide for the claimed subject matter.

Subclass 144 (Fig. 5) is the first subclass to pro-
vide for the claimed subject matter. We now know
that the patent must appear as an original either in
this subclass or in one of its indented subclasses.
We then proceed to the coordinate subclasses in-
dented under subclass 144 and repeat the procedure.
This process is repeated until the ultimate coordinate
subclass is found, in this case subclass 163. The
patent discloses but does not claim the adjustable
line of force (subclass 164). Therefore the patent
may be classified in subclass 164 as an original
patent, since the disclosure may be relied upon for
placement in an ultimate idented subclass.

We have another feature recited in the claims,
and that is the turnplate. So a cross reference copy
should be placed in the subclass that relates to turn-
plates, namely, 166. So far as the additional dis-
closure is concerned, it then becomes the duty of the
classifier to make a subjective determination. This
subject matter is disclosed. Is it valuable for ref-
erence purposes, or not? If, in his estimation, the
subject matter disclosed but not claimed is valuable,
he may place a cross reference in the places pro-
vided for. The automatic control feature would be
cross-referenced into subclasses 43 and its indented
subclasses.

In summing up the rules for placing the original
copy of the patent, as among the 308 classes,
claimed disclosure controls in accordance with a
superiority schedule, the class definitions and the
actual placement of prior patents. As between
coordinate subclasses, the claims control classifica-
tion of the original copy. As between a subclass and
its indented subclasses, the original patent may be
placed on disclosure. Insofar as cross reference
copies are concerned, claimed subject matter must
be cross referenced while disclosed subject matter
may be optionally cross referenced at the discretion
of the examiner.

Figure 3

PATENT

SING
MEANS

DRIVE

CLAIM

A press comprising a top roll and two
base rolls arranged in pyramidal fashion,

the top roll being spring biased toward
the two base rolls thus generating plural
nips,

and a turn plate positioned between the two
base rolls to guide material to be pressed
from the first nip to the second nip.

Figure 4

ADDITIONAL DISCLOSURE

Automatic control of the drive
means

Adjustable line of force

I hope that I have been able to give you some
idea of our problems and the theories on which we
try to base our classifications, and some of the rules
controlling the administration of the system.

In drawing my remarks to a close, I would like
to read a very brief sentence from the classic work
A System of Logic by John Stewart Mill: "There
is no property of objects which may not be taken as
the foundation for a classification, and in our first

attempts we are likely to select for that purpose
properties which are simple, easily conceived, and
perceptible at a first view. But these classifications,
those involving simple and easily conceived things
which are perceptible at a first view are seldom
much adapted to the ends of that classification."
From fifteen years of experience in the classification
group, I believe that I can fairly safely say that our
classification is rarely simple, it is never easily con-

100

144 CONCURRENT PRESSING AND CONVEYING

145 Helix type compressor

146 Right and left helices

147 Overload release (yieldable choke)

148 With valve or choke adjustment

149 Detachable choke

150 Helix interengaging abutment

151 Endless conveyer type

152 With conveyer charging or discharging

means

153 Co-acting presser roll

154 With intermediate backup device

155 Roll type

156 Non-rotary co-acting press element

157 External and internal rolls

158 Intersecting or inclined roll axes

159 Vertical roll axes

160 With backing roll

161 Plural stage or pass

162 Common roll

163 Yieldable

164 Adjustable line of force

165 Differentially

166 With turn plate or roll

167 Chute or conveyer between stages

168 Roll adjustment

169 Yieldable

170 Fluid pressure

171 Spring

172 Roll drive

173 With material handling or guiding

174 Roll strippers

175 Roll bearing guards

176 Roll pairs

Figure 5

ceived, and it is hardly ever perceptible in the first
view. Therefore, they should be magnificent, at
least in the view of John Stewart Mill.

Mr. Green : When I was a member of the examin-
ing force (which goes back to pre-World War II)
my colleagues and I used to speculate quite a bit
about what was behind the decisions of the Classi-
fication Division. You know, did they toss a coin?
Did they turn a card? Did they roll the dice? Now,
over twenty years later I learn that there is a sort of
rude logic in what they do.

Our third speaker is the Honorable L. F. W.
Knight from the United Kingdom. He is Super-
intending Examiner in the United Kingdom Patent
Office, and his extensive career encompasses Cam-
bridge University where he took his degree. He is
a physicist, a chemist and a barrister. His record in
the Patent Office spans the period from 1930 to
1961 with time out for some war activity in the
Ministry of Transport. Currently, he is in charge
of all classifications, mechanization and the examin-
ing divisions. He was Chairman of the United
Kingdom's Study Group on classification, and this
afternoon he will describe to us the report of that
group. The Honorable L. F. W. Knight.

Mr. Knight: Mr. Moderator, Ladies and Gentle-
men: Almost exactly two years ago there was set
up in the United Kingdom Patent Office a Study
Group to inquire into and report upon the classifi-
cation system and problems arising from it. Of that
Study Group, as your Moderator has told you, I had
the honor to be chairman. The terms of reference
of this Group were very wide indeed, including
power to recommend changes of any sort in the sys-
tem, and to advise on long term requirements for
proper control of clasification. The Study Group
felt that it must look into other systems in addition
to our own, and that its province included filing sys-
tems and aids to searching as well as classification
proper, but that it did not extend to mechanical
retrieval systems per se, although these could not be
ignored since their use clearly affects the type of
classification required.

Whilst we realized, of course, that there was a
public interest as well as an examiner interest in the
classification system, we concluded, and I quote from
paragraph 123 of our report: "So far as examiners
are concerned, the purpose of the classification sys-
tem is to enable them to make as quickly, econom-
ically, and reliably as possible, the investigations
prescribed by Section 7 and 8 of the Patents Act,
1949." With Section 8, which relates to prior claim-
ing, I need not bother you, but I must, I think,
quote the relevant part of Section 7. I quote: "The

examiner shall make investigation for the

purpose of ascertaining whether the invention, so
far as claimed in any claim of the complete specifica-
tion, has been published before the date of filing
of the applicant's complete specification, in any
specification filed in pursuance of an application for
a patent made in the United Kingdom and dated
within fifty years next before that date." Section 7
( 2 ) extends the scope of the investigation at the
Comptroller's discretion to any other document
excepting only patent specifications of any country
which are more than fifty years of age, or any
abridgement or official extract officially published
of such specifications. It will be noted, then, that
the scope of our search differs from that of some
other countries in certain respects. First of all, we
search only for what is claimed. We rely absolutely
on the claims to determine the scope of the invention,
although the descriptive part of the specification
may be called in aid to interpret the claims. Sec-
ondly, the search we make under Section 7 is purely
a novelty search. Thirdly, our search is statutorily
limited as far as patent specifications of any country

101

whatsoever are concerned to a period of fifty years
preceding the date of the specification. Fourthly,
we are only bound, statutorily, to search United
Kingdom patent specifications. Any extension of
that search to specifications of other countries or to
literature generally is at the discretion of the
Comptroller.

It will also be appreciated that for over fifty
years we have had a rather more elaborate general
filing system than, I think, any other country, a
system which might even be regarded as an embryo
mechanical searching system. And as a result of
this, and of the limited objectives of Section 7 of the
Act, searching does not occupy even now a very
long time in the United Kingdom Patent Office.
The average throughout the Office is only about
three hours per application. On the other hand,
despite the comparative shortness of that period,
we think it is much too long and one of the objec-
tives of the Study Group was to see if we could
reduce that three hours to one hour.

With this preamble over, I now propose to turn
to the Report itself. The actual report is a docu-
ment of 1 29 foolscap pages. I am not permitted
to give you copies of it, and in any case I am quite
sure you would not want them. It was signed
and presented on the 9th of August 1961, but it
has not yet been published. Some of its recom-
mendations have, however, been accepted, whilst
others are still sub judice. None as yet have been
turned down completely. Despite the fact of non-
publication, I am, however, authorized to summa-
rize at this meeting the work of the Study Group
and those recommendations embodied in the Report
which have anv bearing on the subject matter of this
Workshop.

What I propose to do is to read or summarize
parts of the earlier sections of the Report, which are
factual statements of present positions, since without
those you might not appreciate the point of some of
the recommendations for change, and then to tell
you the gist of our main recommendations insofar
as they relate to our present topic. There are many
other sections of the Report that I would like to
deal with, but this would take far too long a time.
I will merely mention them in passing. I would,
however, ask you to bear with me if I diverge to
some extent from classification pure and simple, since
I am primarily reviewing the Report and not
describing to you the United Kingdom classification
system.

The introductory section of the Report and that
part of the first main section dealing with the his-
tory of classification in the United Kingdom, and
with the Comptroller's obligations and powers I can
pass over. But I must go into the present classifica-
tion system as described in paragraph 19 onwards
of the Report if you are to appreciate the relevance
nf our recommendations. I therefore ask you to
bear with me while I read a rather lengthy passage
from the Report itself:

"The classification system comprises the subdivi-
sion of the whole field of patentable invention into
a large number of pigeonholes relating to more or
less definite subjects, and the rules, written and un-
written, by which a specification is indexed, i.e., its
subject matter is allocated to the pigeonholes. The
physical expression of the system is the Key, the Ref-
erence Index, and Office Practice," corresponding
very much to the three volumes in the American
System, I believe. — "The products of the system,
and matters closely related to and dependent on it,
are the search files, abridgement volumes, the sub-
ject matter index in the Official Journal (Patents),
classification codemarks on printed specifications,
and the punched card system for recording and re-
producing the index and allotment of specifications.

"The Key is essentially a list of divisions of sub-
ject matter which may form the subject of invention.
The primary divisions are headings denoting more
or less definite fields of invention, and these head-
ings, over 1400 in number, are collected into 270
numbered classes, themselves assembled into 44
numbered groups for the purposes of abridgement
publication.

"Almost all headings are followed by one or more
of the following items: qualifying notes in paren-
thesis, excepting notes, italic notes, and cross-
references directing attention to other headings." —
I hope you will remember that our term "cross-
reference" in the United Kingdom has very little
connection with the term as used in the United
States Patent Office. It indicates in the United
Kingdom the place where you will find that subject
matter. It is an entry in the Key itself saying that
such and such subject matter is taken out of this
heading and put somewhere else. That is the mean-
ing that we give to the term "cross-reference". —
"Most headings are also followed by subdivisions
within the scope of the heading and many of these
subdivisions have subsidiary cross-references and
are further subdivided, often in successive steps to
increasing degrees of fineness. Each level of sub-

102

division is shown by indentation, and its degree is
shown by the number of dots which precede the
subdivision." — We have rows of dots correspond-
ing roughly to your vertical lines. We prefer hori-
zontal dots for some reason or another. — "Each
undivided heading and each ultimate subdivision
of subdivision of subdivided headings is labeled in
the margin with a clasification code-mark." — That
is another difference from your classification sys-
tem. It is only the ultimate subdivision which is
numbered. If a subdivision has further subdivisions
then the higher subdivision is not numbered.

"The groups may be regarded as forming the
largest units of classification. Their sole practical
value, however, is in classifying abridgements for
publication in separate volumes. They have no
other real classification function; for instance, no
use is made of them as a stage in deciding where
any matter should be indexed. The scope of each
is undefined and is merely the sum of the scopes of
the classes which it comprises.

"The 270 classes are the next largest units of
classification. They, too, now have no real classi-
fication function, and their main uses are in the
distribution of search materials and specifications
for examination to the examining groups and in
the arrangement of recently published specifications
in the library. The scope of each is, again unde-
fined and is merely the sum of the scopes of the
headings which it comprises. Their number order
has no practical significance and arises from an
original alphabetical order.

"The 1,400-odd headings are the largest units
which have any real classification significance. The
majority relate to particular fields of use such as
specific industries (e.g. Pottery, Manufacture of)
or machines (e.g., Typewriters etc.) which are
distinct from one another, but there are a number
(e.g. Bearings etc.) which relate to matter having
applications in the fields of other headings. The two
sorts are frequently called by examiners, rather
loosely, 'application headings' and 'general head-
ings'. These terms are relative rather than
absolute. . . .

"Each heading is intended to have a more or
less difinite scope, but the determination of this is
a highly sophisticated matter. . . . The breadth of
headings varies widely from one to another, and
some which are broad in their terms have excepting
notes and cross-references which exclude most of
the matter that they would otherwise cover, and
consequently comprise only residual matter.

"Italic notes which have been increasingly used
in recent years are mostly intended to elucidate in
one way or another the scope of the heading, by
defining or explaining it, or saying what it includes
or excludes, or some combination of these things.
A few of them explain rules for indexing under the
headings where they appear, or say what is repeated
under another heading." In other words, the head-
ing in the Key is explained in the Key. There is
not a separate explanatory booklet.

"Cross references are of two sorts. One sort indi-
cate where accessories, details, and other matter
cognate to a heading are indexed. The other sort
exclude specified subjects which are within the
terms of the heading or subdivision and transfer the
matter specified in the first part of the reference to
the heading or subdivision specified in the second
part of the reference." Both are excluding — they
exclude subject matter from the original heading,
but one excludes subject matter we would expect to
find there; the other excludes the accessories and
details which you might not expect to find.

"Each subdivision, like a heading, is intended to
have a more or less definite scope.

"The scope of a heading is initially indicated by
its terms in conjunction with any italic notes, ex-
cepting-notes, and cross-references appearing under
it. The excepting-notes and cross-references as
stated above indicate matter that is not indexed
under the heading under which they appear. The
effective scope of the headings, so cut down, is fur-
ther affected by cross references under other head-
ings (no indication of these is given under the re-
ceiving heading) and by certain rules and practices
in indexing.

"The first of these rules is to treat the heading as
though it were restricted by an indefinite number
of unwritten but implied references concerned with
features which, though given a particular applica-
tion in a particular specification, are regarded as of
such an obviously general character that they could
not appropriately be indexed under any heading
relating to a particular field of use (33 examples of
these are given in an explanatory note which
prefaces the Key)." To give you an example, one
must always assume an implied reference to
'alloys'.

"The second rule is not to allow the mere form
of wording used in the heading to exclude matters
closely related to its substance, unless of course there
is another heading which clearly takes such matter.
Thus, subject to the proviso just mentioned, a head-

645803 0-63-3

103

ing restricted in form to processes is regarded as also
covering apparatus and materials used in perform-
ing the process, and a heading worded to cover only
apparatus or substances covers also processes and
apparatus used in making or treating the apparatus
or substance.

"The third rule is to interpret the wording of the
heading not literally, but primarily as expressing a
central idea around which similar subject matter
may be grouped.

"In addition to these rules, the following seem
to be the main rules used in indexing: (1) Every-
thing that is described in a specification that is likely
to be of interest to a future searcher, whether it is
strictly relevant to the invention claimed or not,
should be indexed. (b) In deciding what is likely
to be of interest the practice is generally as follows :
(i) If the matter is well known and represented in
the search-files of the heading and is at the same
time not closely related to the claims of the specifica-
tion being indexed, it is not per se of sufficient in-
terest to that heading, (ii) The mere mention of
a particular detail or element of an apparatus or
process does not by itself (with a few exceptions)
make it of sufficient interest to a heading to warrant
indexing under it. (iii) Under many headings
which cover the apparatus or process as a whole, a
practice similar to (i) and (ii) is applied in decid-
ing whether to index a detail or element, (iv)
There are numerous exceptions to (iii) ; their extent
is not certain, and the areas in which they occur and
the forms they take are rarely indicated in the key.
Under some headings any combination of elements,
considered as one whole, is met, not completely in
one particular subdivision providing for such a com-
bination, but under the subdivisions for the various
elements, even if each element is in itself well known
and perhaps not even described. This is done to
facilitate searching by using a combination of sub-
divisions representing such elements (combination-
clip searching) which is, to some extent, the fore-
runner of mechanical searching systems. In other
headings, features without novelty in themselves
are indexed in subdivisions at a certain level only if
there is novel matter of interest to an associated sub-
division on the same level, (c) Any particular
feature of interest, whether it relates to a detail or a
complete process or apparatus, is usually indexed
under one heading only, although in some cases it
may be indexed from different points of view in two
or more headings. In any one heading it may be
indexed under one subdivision only, or under more

than one subdivision if the heading has several sets
of subdivisions corresponding to different aspects of
the subject matter (e.g. shape, material, purpose).
(d) If there is no clear home for some matter, the
nearest suitable one is chosen, or the Key is modi-
fied or amplified to take account of the matter in
question, (e) A machine or apparatus described
as a whole should be indexed under a heading or
subdivision which denotes the kind or type of
machine or apparatus. No similar rule explicitly
exists for processes. Even for machines or appara-
tuses it is not uncommon for there to be
no heading which covers the whole machine
or apparatus, and recourse ■ is then had to
indexing under headings covering the parts, (f)
In a machine or apparatus described as a whole,
details which are described and which have an in-
terest beyond their mere application should be met
under detail subdivisions or, when appropriate,
under other, general, headings, (g) An improved
article, whether or not it results from the use of a
process, apparatus, or machine, indexable as such,
is indexed under the heading relating to such
article."

That is the greater part of the section of our Re-
port setting out the present system. There are
another seven or eight paragraphs, but they deal
with such small details that I do not think I need
bother you with them. I am sorry I have had to
read all this in extenso, but without it I do not think
you would understand the purpose behind our
recommendations.

I am now going to skip some 220 paragraphs of
the Report in about thirty seconds, I hope. We
next went on to review other systems, notably those
of Germany, the Netherlands, and the United
States, none of which reviews I will, or scarcely dare
in this company, read. We also reviewed the Inter-
national Patent Classification which my friend,
Rubach, has already dealt with most competently
this afternoon. We also reviewed the history and*
present work of our classification sections, the ques-
tion of present day responsibility, and the extent of!
reclassification undertaken in recent years. That
was the extent of Part I of our Report.

It was followed in Part II by a detailed, critical
appraisal of the system as it stands at present, its*
characteristic features, the conflicting requirements
of the public and the examiner, the complexity and
illogicality of the present system, the apparently con-
tradictory needs of stability and flexibility, the
structure of the Key and of the Reference Index, the

104

impact of mechanical searching, filing systems and
other aids to searching, unofficial classification
schemes, problems of the scheme of headings, prob-
lems of individual headings, problems of subdivi-
sions, problems relating to particular types of head-
ings, such as general and application headings,
compositions, articles and materials, and so on,
specific problems between headings, and finally,
internal problems of particular headings.

Then we come to the third part of the Report,
which embodies our recommendations and a good
deal of that I propose to read to you. In paragraph
267 we say : "After careful consideration of possible
alternatives, we do not recommend any major de-
parture from the method at present in use," the
present method being to define such features as pre-
cisely as possible by descriptive wording.

The next recommendation appropriate to this
Workshop is that every item of subject matter in a
patent specification should be indexed. "We con-
sider it essential that this practice should remain
unchanged. . . . Claims can be, and frequently
are, made for combinations of features, some or all
of which may be known per se, and for minor im-
provements in articles, apparatus and processes of
known type, and in these circumstances failure to
index features, because they are considered to be
incidental to the invention, or do not relate to the
apparent inventive step, would clearly reduce the
effectiveness of the search."

The next one : "We recommend the provision of
a logical or substantially logical superstructure to the
scheme of headings in place of the present abridge-
ment groups and classes, which have no real classi-
fying function. This would enable the correct home
for any particular item of subject matter to be de-
termined by reasoning process, and would facilitate
amendment of the Key to provide for new matter.
. . . We also recommend that in the interests of
simplicity, comprehensibility, and economy, the
number of headings should be drastically reduced,
e.g., to about 450.

"We are also of the opinion that the system should
have some relationship with the I.P.C." — The
International Patent Classification — "As the latter
is not entirely logical from our point of view, this
raises the difficult question as to whether or not logi-
cality should be sacrificed in order to secure a rea-
sonably close relationship between the two systems.
We have devoted much careful thought" — or so it
says in the Report — "to this question and have con-
cluded that the examiners' system should be brought

into close agreement with the I.P.C. (at least at
higher levels of classification) even if this involves
some departure from logicality in the proposed
superstructure to the scheme of headings.

"Accordingly, we recommend that the first stage
of classification should correspond to the eight sec-
tions A-H of the I.P.C, the second stage should be
represented by 41 divisions, each corresponding
approximately to a single class or group of adjacent
classes of the I.P.C, and the third stage should com-
prise about 450 headings arranged as far as possible
logically within these divisions. The scheme of
headings should be produced initially by condensa-
tion of the present scheme with the minimum dis-
turbance of existing major headings. The proposed
41 divisions are listed in appendix G to the
Report. . . .

"The scheme of headings should be made as com-
prehensive as possible to avoid undefined gaps and
to enable subject matter to be indexed freely from
different aspects. In the past, much valuable index-
ing information has been deficient in these respects,
and we are strongly of the opinion that this should
be avoided in the future, particularly in view of
the possible advent of mechanical searching on a
large scale. The simplified scheme of headings,
detailed in Appendix H, is not comprehensive,
having been produced by condensation of the exist-
ing 1400-odd headings. We regard it as a possible
step toward an ultimate goal. . . . Subject matter
should not be indexed in a specific heading which
does not strictly cover it unless the heading in
question is amended immediately. If such amend-
ment is inconvenient at the time, the subject matter
in question should be temporarily accommodated
in an associated miscellaneous heading which should
be reviewed at frequent intervals.

"Pending development of the I.P.C. from its
present embryo form, we are unable to predict the
extent to which it might ultimately be used by the
examiner for search purposes, but we doubt whether
it will even be developed in sufficient detail to
provide by itself a satisfactory substitute for our
present type of system. However, if a class or sub-
class of the I.P.C. provides a suitable unit which
could be adopted as a heading, we recommend that
it should be adopted when and where convenient,
and if the I.P.C. provides a suitable basis for the
subdivision of the heading, we suggest that it should
be made-use of when that heading is re-classified,
unless there are compelling reasons for adopting a
different course. We express the hope that, by

105

development on these lines, the structure of the
system used by the examiner will gradually approach
the I.P.C. The ultimate objective might possibly
be a system which incorporates all or most of the
I.P.C. subdivisions, but is modified or amplified in
detail as necessary to suit our particular require-
ments."

There then follow in the Report the more de-
tailed recommendations regarding the form and
interpretation of headings and subdivisions, and the
recommendations relating to specific headings.
These, I am not going into, but I would direct your
attention to one or two of our later recommenda-
tions. In paragraph 314: "Whenever major re-
classification of a heading is proposed, we
recommend that the possibility of utilizing punched-
card or equivalent records for search purposes
should be carefully explored and if considered ex-
pedient, the new classification should be made
suitable for their employment.

"Steps should be taken to educate examiners in
the use of Batten and similar systems and to encour-
age them to adopt a less conservative attitude to-
wards nontraditional methods of searching." I do
not know whether you get such examiner resistance
in the U.S. Office: We do in the British Office.

"Pending further development of mechanical
searching within the Office, we recommend that re-
search should be carried out to ascertain whether
the existing cards and clips could be modified to
enable a greater amount of indexed information
to be presented to the searching examiner in a con-
venient manner.

"Examiners should be trained to appreciate the
various ways in which search files can be organized
to enable search time to be reduced, or searches to
be made more effectively . . . and should be pro-
vided with more detailed information on the use
of colored clips and the employment of various other
searching aids."

Another recommendation to which I would like
to draw your attention, although it has strictly noth-
ing very much to do with this meeting, is this:
"We recommend that a training scheme should be
instituted whereby selected examiners should be
attached to Classification Section for a period of
(say) 18 to 24 months to receive instruction and at
the same time to assist in the general work of the
section." We suggested this training should be
carried out when an examiner has been in the Office
from five to ten years.

I won't keep you much longer, but I would like
to read those parts of the Report which deal with
our proposals as regards the duties and powers of
the Classification Section; I do not think it will
take me very long. "We recommend that Classifi-
cation Section should continue to check the correct-
ness and completeness of index and search, at least
down to heading level, of all specifications at the
acceptance stage, and that steps should be taken
to investigate how far this procedure needs tighten-
ing up. In our view, the check will be particularly
necessary if and when the changes resulting from
our previous recommendations are put into effect.

"We also recommend that Classification Section
should make a similar check of the I.P.C. Index
when the latter is done by group examiners.

"We recommend that the section should have
power (a) to determine the scope and wording of
each heading in the Key and of the boundaries be-
tween such headings, (b) to decide the headings
in which any specifications is to be indexed and
searched, (c) to direct that a particular specifica-
tion be examined in a group which is not respon-
sible for the primary search files, e.g. where special-
ized technical knowledge is required, (d) to issue
to the examining staff, instructions on classification,
searching and indexing and to resolve disputes on
their interpretation, (e) to approve of any pro-
posed internal classification of a heading; major
changes should be subject to approval before the
work is started, (f) to direct that internal reclassi-
fication of headings should be carried out, (g) to
direct the lay-out of search files and the use of
mechanical search systems, (h) to allocate classi-
fication time within an overall allowance of time
for the whole examining staff, (i) to examine or
search specifications, e.g. if a new classification prob-
lem is presented or it is desired to ascertain the state
of classification in any particular field, (j) to re-
quire the furnishing of information for classifica-
tion purposes, and in particular to direct that
searches should be made for these purposes.

"We also recommend that Classification Section
should have its duties extended to include the fol-
lowing: to plan and supervise the transition from
the existing classification system to a modified sys-
tem proposed in this report ; to discover and investi-
gate defects of the system, in particular, by (i)
periodical examination of statistics relating to the
number of indexes and searches in each heading,
with a view to abolishing or revising headings in
which both are low, or in which there is a wide dis-

106

parity between the number of indexes and searches,
(ii) the maintenance of records of subject matter
which cannot be adequately indexed under the ex-
isting scheme of headings, e.g., inventions which can
only be indexed under a miscellaneous heading, in-
ventions which are capable of general application or
use which can only be indexed under application
headings and specified applications of inventions for
which there is no provision in the Key, and (iii)
periodical preparation and examination of statistics
relating to the average search times in the various
headings; to initiate proposals to remove, as far as
possible, the defects so discovered, and to direct that
a heading should be reclassified if the average search
time is considered to be unnecessarily long; to coop-
erate with examiners in the implementation of re-
classification schemes." In the past it has been the
practice for the examiner in charge of the class to
carry out the actual classification, although subject
to advice from Classification Section; "to revise and
sanction all proposals for amendment of the Key;
to advise on layout of examiners' search files and the
use of colored clips and other searching aids; to
maintain an up-to-date Reference Index; to control
and develop mechanical searching; to advise on the
administrative setup of examining groups," and one
more recommendation. But I pause here for a
brief moment before reading this last recommenda-
tion on the duties of Classification Section, since it
is particularly appropriate at this very meeting that
we find this prophetic recommendation imple-
mented — I quote : One of the duties is " — to consult
with other Patent Offices and with other classifying
bodies on problems of classification and the use of
searching aids."

Mr. Moderator, Ladies and Gentlemen, I thank
you.

Mr. Green: That was a very rewarding presenta-
tion. I think we now have the opportunity for the
coffee break which will give you a second opportun-
ity to think up some very difficult questions for our
three speakers when we return. I believe we ought
to be able to get back here in twenty minutes with-
out any difficulty.

A short recess was called by the Moderator.

Mr. Green: I have been told that the conversations
that were going on in the Delegates' Lounge dur-
ing the intermission were as rewarding as any other
part of the Session, and I am confident that this is
true. So, perhaps we might aim to conclude our
formal discusion in about half an hour, and for

those who wish to carry forward, the quarters are
here and I am sure you will be seeing a lot of each
other for the next couple of days. The ground
rules are relatively simple. As you know, these pro-
ceedings are being recorded. Therefore, it would
be desirable if you would speak up, ask your ques-
tion, make your observation and identify yourself.
Now, the Program says that the round table dis-
cussion is going to be limited to two topics. Of
course, that is not so at all. Let's talk about any-
thing that is pertinent.

Mr. Rot kin : This is not in the nature of a question
as much as bewilderment. The speakers today have
touched upon the classification systems that exist, or
that they would like to see exist. But no one, as yet,
has spoken about how such a classification system
is created. In line with the first topic for round
table discussion, I think it would help, if we were
to give some thought and vocalize on the techniques
or mental processes that go into the creation of a
classification scheme. Not how one encodes or al-
lots the pigeonhole notation on a patent into an
existing system, because if we were to find that in
fact we have the same mental processes in creating
our classification system, it might be easier to get
together in a somewhat universal classification
system.

Mr. Newman: One of the most useful methods of
classifying is to take like things, discover their dif-
ferences, and use the differences as items in a hier-
archical classification system. This actually, in my
personal belief, is what the Classification Division
of the U.S. Patent Office does, when it synthesizes
its classification schedules. You make piles of doc-
uments concerning similar things. You study them
and find out what their differences are. You then
list these differences as elements in the classifica-
tion schedule. This is one of the ways in which one
can develop a highly usable classification system,
because it is these differences which continue to
grow, and you must make place for the thing that
is going to come into being tomorrow.

Mr. Green: Does any one of our vistors from abroad
wish to comment?

Dr. de Haan: It is with the greatest interest that I
listened to the addresses of Mr. Rubach, Mr. Vin-
cent, and Mr. Knight. I see that, more or less,
they have defended a certain classification system
with which they are acquainted. The first speaker
asked, why it would not be possible to come to a
coordination Listening and thinking about the

107

remarks made by the speakers, it reminded me of
Shakespeare who said somewhere that there is noth-
ing either good or bad but thinking makes it so.
It strikes me always that only very higly intelligent
gentlemen deal with classification. I can under-
stand that having thought for many years about their
own classification systems they each think that their
own sysem is excellent. But on the other side, it
might be that these same intelligent gentlemen con-
sidering other classification systems which depart
from their point of view, might think that these
classification systems are partially or totally inade-
quate to their needs. I may, too, say that they
might think these systems bad. And so I fear, al-
though the U.S. system, the U.K. system, the Ger-
man system, and the I.P.C. in themselves are very
good systems, with which examiners who have
studied the systems can work, it would be very diffi-
cut to work them together. And I am not very
optimistic that the existing classification systems for
hand searching can ever be brought together.
Throwing away your own classification system and
replacing it by another system gives such a lot of
work and expense that for practical reasons only
it might be impossible. But we have heard that with
hand search classification it becomes more and more
difficult to search the enormous number of applica-
tions and the enormous amount of documentation,
and that in one way or another, we must come to
mechanical classification work, which is also one
of the most important topics of this Workshop.

I should like to draw your attention to the fact
that if we introduce the mechanical system for
searching, we have to throw away our existing
searching systems. And if we try to work out a
mechanical system which, as I feel, is a very, very
difficult thing, and which will take many years, it
can only be accomplished by very intelligent classi-
fication gentlemen in the various Patent Offices
working together and thinking together, in the same
way. What it may be is a second question. But
only if they think together, can we get one mechan-
ical system that is the same for all our Patent offices,
and which we will all think is really a good system.
We would have the great advantage that communi-
cation of information among the Patent Offices and
documentation of industry in our various countries,
could be done on the same basis and that we could
economize to a great extent on expenses and labor.

Mr. Green: It seems to me that the logic of what
has just been said is obvious. When you enter into
machine language, your vested interests in specific

108

classification systems should merge into a new sys-
tem. If I might be permitted I would like to tell
you of a little work that is going on in the United
States now, not in the Patent field, but directed
toward exactly what you described. In the Fed-
eral government we have a number of documenta-
tion services, one in the Defense Department, one
in the Atomic Energy Commission, one in the new
Space Agency, and one in the Department of Com-
merce under my jurisdiction. Four agencies are
going ahead to mechanize in a compatible fashion
so that we can exchange tapes, so that we can merge
our effort, so that we can divide the operation. We
find that this is working in a most effective fashion
in the early stages. And it just seems natural logic.
Well, if I may be permitted this intrusion, any other
observations?

Mr. U eland : In Norway the examiners have to deal
with three filing systems, the German system that
we follow, the American system, and the Dutch
system. All British and French patents are classi-
fied after the Dutch system. It would simplify the
examination of novelty if all patents were classified
according to one classification scheme. I under-
stand, of course, that this is difficult because in most
European countries the inventions are classified
according to technical use of the invention, while in
the U.S.A. most of the patents, I deem, are classified
on the object and, I suppose, of the object's end use.
In a few years I presume all the European patents
will be classified also according to the International
Classification system, and in that way retrieval of all
patents will be easier. I should like to know if it
would not be possible to glean from the U.S. pat-
ents, in addition to the U.S. class, the International
Class designations. Because of the great number
of patents granted here, the U.S. material has been,
and will be, an important searching material, but it
is very difficult for us to use it. The U.S. Patent
Office very often changes its classes. Some classes
disappear. Some new classes are formed and pat-
ents are transferred from one class to another. It
is extremely difficult to retrieve the patents after all
the changes are made. An example — Class 13,
Electric Furnaces. A number of patents from that
class are spread over fifteen new classes. I suppose
that after a complete International Classification
system is introduced, there will be a few changes for
some time to come. By printing the International
Classification on the U.S. patent, some stability in
the classification will be obtained and the retrieval

will be easier. It will be of the greatest importance
for international patent cooperation.

Mr. Green: I think possibly the answer is that this
is something for the Commissioner of Patents to
study, and I doubt that an answer can be given this
afternoon.

Mr. Rubach: But this question was often placed be-
fore us. The industry and the agents come and ask
if we could not give them a correspondence list be-
tween the German classification and the classifica-
tion of the U.S. Patent Office. We have tried a lot
of times. It is absolutely impossible to make a
correspondence list because of the way in which we
classify, even in the International Classification sys-
tem and the way in which the Patent Office of the
U.S. classifies. I must say I would appreciate it
very much if this would be possible. But I might
also say I can not imagine that the U.S. Patent Of-
fice could do that.

Mr. Green: My own impression is that the only
people who think it can be done today are the bright
young men who sell computers.

Mr. van Waasbergen: I would strongly underline
what Mr. Ueland has said, because the classification
of United States patents in the diverse European
Patent Offices is a big problem. The United States
Patent Office patents arrive at The Hague and in
Munich, and in Stockholm, and in many other
places, and if you want to have a consistent classi-
fication system at all these different places, reclassi-
fication must be done. That is to say, at several
places, exactly the same work is done. It would be
much better if the work were done once in the
United States Patent Office itself. If the United
States Patent Office could print the index of the
International Classification that would be a very
big advantage for all the other offices which are
going to use the International system.

Then, I would like to add a few words about the
possibility of a concordance. You said, Mr. Green,
that perhaps with a computer it could be done. I
do not think this is true. If you consider one sys-
tem of classification, for instance the International
system, you could express it in a graphic way. I
say, you divide the arts in horizontal rows. But if
you have, for instance, the United States classifica-
tion system, you do not divide it horizontally, you
divide it vertically. And it is quite clear that it is
impossible to find a horizontal row which corre-
sponds to a vertical column. So, the concordance,

in principle, in my opinion, is impossible. The only
way to get the indexes of another classification sys-
tem is that we have the men who know it and want
it, and the men of the other system classify it and
put the indexes on the paper.

Mr. Green: My comment was facetious. I was in-
tending to say that only the people who sell equip-
ment will promise you that they can do this, and not
the people who do it.

Mr. Fairthorne: As one who has been a computer
or computor, for over a quarter of a century, I am
in the unusual position of defending computing
machinery, because I generally attack very vigor-
ously on this thing. I would say there is very good
hope, indeed, in the not too far future, for the
use of computers to allow a lingua-franca interna-
tional form of classification for use, subsidiary to
what any particular national patent office or indeed
any national library uses. The problems of making
concordances and so forth are very considerable.
And what you call a concordance in this particular
connection is something very much more elaborate,
of course, than anything that the Jewish, for in-
stance, make a concordance in the Bible, of which
we have had some practice, because if my memory
serves me right, one of the earliest belongs to the
sixteenth century in the book of Rabbi Isaac Nathan.
Well, we need something more elaborate than that.
But, although it hurts me to say so, the contem-
porary machinery can do quite elaborate work of
turning things of one structure into another, pro-
vided they have enough capacity, enough storage.
Current computers, indeed for the next two years,
will not have anything like the amount of rapid
access storage for what we have in mind. In the
next five years I suggest they might. It is not like
waiting for an artificial intelligence to come out of
a tin box or anything. It is a question of simply
how big they are going to get. And I would suggest,
therefore (again, it goes against the grain) that
people forget about computers in searching here
(which in my opinion and some others, is the least
important of all the applications of machinery in
this field ) and consider its use for converting classi-
fications, revising classifications and so forth as
bits and pieces come in. For that sort of work even
contemporary machinery can be quite useful, and
I should say that in three or four years will be a
great deal more than quite useful. It will allow any
particular classification to be represented, in stand-
ard form, should I say, whatever people care to use

109

in their own countries. This, admittedly, is specu-
lation, but I do not think it is unreasonable.

Mr. Green: Thank you. That is a very useful
observation. Any further discussion?

Mr. Jeffery: As a representative of quite a small
patent office, it is with some trepidation of course,
that I dare to open my mouth here, amongst the
computers and classificationists. But we have had
some experience with two classifications, the Aus-
tralian and the International. And we find after
two years, that it is impossible to have a concordance
in the sense that, if this is the Australian classifica-
tion, this must be the International classification.
There is not a one-to-one relationship. But let us
get right back to what is the purpose of classification.
It is to help searchers in the future. Now, a
searcher is searching for some specific subject.
Could he not have an international word list, or
subject list, whereby, if he ran down the column and
came to the subject he was searching, which might
be, say, alloys, magnesium, aluminum, and running
across he sees the various classes under the different
countries. You see, it is back to front. He ap-
proaches with a word and he finds the class. So I
throw this one in for general discussion.

Mr. Newman: We have a possibility here of getting
a partial concordance. The U.S. examiners classify
the foreign patents in their own classification scheme
as they come in, and we have at the present time in
the U.S. Patent Office a Hollerith file of foreign
patents and their U.S. classification. Now, it seems
to me that somebody could get hold of these decks
of cards and resort them, other than by country.
One could have at least, a list of those foreign
classifications in which the majority of patents were
coordinated and also a list of where a U.S. patent
classification or a foreign patent classification did
not overlap.

(Further. discussion brought out the fact that this
classification was uncontrolled, since it was applied
by individual examiners, and was therefore too
variable for serious consideration. )

Mr. Rot kin: I might comment, Mr. Newman, that
I tried that on a very, very small scale when we first
got the printed copy of the International Classifica-
tion scheme. I could not find enough coordination
to say that there was a majority of anything. In
other words, the divergence of the numbers, in the
small area I looked in, made me despair. I would
like to see it tested a bit more rigorously and scien-

110

tifically. I would like to add just one further
thought. The suggestion of putting the Interna-
tional Classification on the American patent, I
think, is a matter of manpower, or, to be more
fundamental, dollars and cents. It would take
some three or four men, versed in the various
mechanical, chemical, and electrical arts, to actually
do that encoding, because it is strictly a question of
encoding. There is no creative work involved.
You have to be familiar with the classification, that
is, the International Classification, and know
enough about the technology to determine where it
fits into that classification. The problem, though,
I think goes beyond that. A single notation would
show up even more alarmingly the difference in the
basis on which patents are issued. In the European
system, particularly in Germany and in Switzerland,
you have the so-called single entity of thought with-
out embellishments in a patent, as distinguished
from the American patent which has a terrific
variety contained in the claims of the same patent.
If you did not provide for additional copies, each
one of which had its own separate encoding for
every variation and for every claim, you would lose
most of the disclosure. Now, the additional copies
would compound terrifically the time and the ex-
pense involved in putting the International Classi-
fication on these copies.

Mr. Green: Admitting that this is a problem of a
great deal of complexity and difficulty, if this many
representatives of foreign patent offices think that it
is a serious problem, I believe that this conference
should call it to the attention of the Commissioner
of Patents, and ask him to seriously study it and see
what can be done.

Mr. Oredsson: Listening to the discussion, I got an
idea similar to Mr. Newman. There are a lot of
applications, priority applications, filed in a lot of
different countries. Let us take, for instance, in
Germany and in the U.S.A. Would it not be pos-
sible to identify, by, for instance, the titles and the
inventor, the patents which are granted on these
applications, and just try to find out where they are
classified according to the German system and the
U.S. system, and in that way perhaps it would be
possible to work out some sort of concordance be-
tween the two systems.

Mr. Green: We will add this suggestion in the
recommendation that we make to the Commissioner
of Patents.

Dr. de Haan: Mr. Moderator, as to the indexing of
U.S. patents with the International symbol, I think
that we Europeans, who are beginning to use the
International Patent Classification, must not bother
the American Commissioner with the task of putting
these symbols on their own patents, for two reasons.
In Europe, in my office, and, I think, also in the
German office, we have been classifying all Amer-
ican patents into our own patent system. I think
if we have a common classification, we should come
together and divide the task and ourselves put the
I.P.C., the International Patent Classification, on
the U.S. patents. Also for a practical reason, be-
cause (and I heard it also from one of the last
speakers) for a classifier who is not accustomed to a
classification system, it is much more difficult to put
these symbols of foreign classification systems on the
U.S. patents than it is for a classifier who really
uses the International Patent Classification. So for
the question of economy, it is far better that we do
it ourselves than that we ask the U.S. Patent Office
to do it.

Mr. Rubach: I would only point out that even if
you have an American specification and a German
specification that may be a branch specification, you
might go back to one particular inventor and one
invention. But if you see what is left at the end
you will see that what we call the invention, is abso-
lutely different in the other country. Therefore, it
is of no use to make comparisons.

Mr. Oredsson: I do not see that it makes any differ-
ence what the invention is in this case, because now,
I think we are talking about documentation, a sort
of concordance. And the disclosure in the specifi-
cation, I think, would be rather similar in both
countries.

Mr. Green: Do we need to pursue this further? Or
can we go on to something else?

Mr. Oredsson: Let me just add another thing. By
picking up this sort of duplication in different coun-
tries, you could perhaps withdraw some of the pat-
ents in the source file, and you would reduce it to
rather a good extent.

Mr. Green: Are there any other observations on this
point?

Mr. Pfeffer: It seems to me that if we are getting
into a discussion of concordances, that we are tak-
ing the easy way out here, and overlooking the main
issue. It would seem to me that the gentlemen from
Norway, Germany, and the Netherlands, who have

indicated the desire to have the International Pat-
ent Classification put on the U.S. patents, are in a
subtle sort of way indicating a desire for some sort
of standardized international classification. I be-
lieve that this should be one of the real issues that
we ought to be discussing here. In listening to the
descriptions of the various systems that were dis-
cussed today, I found some basis for some work in
the future. It appears that they are not completely
at odds with each other. Whether you start from
a system based upon use and develop your general
classes afterwards, or whether you start with the
general classes of function and develop special uses
later on, we eventually will arrive at the same point.
So there is quite a measure of similarity here, and it
seems that this would form a basis for some future
activity in trying to establish some real international
standards.

Mr. Jancin: A thought occurred to me during Mr.
Knight's talk in connection with his statement, that
a search is limited to about four hours, if I remem-
ber, and that they are trying to get it down to one
hour, and that their search by statute is limited to
U.K. patents, and that it may be extended by the
Comptroller. Isn't this due in a great part to the
practice in the U.K. of oppositions and to the main
claim procedure practice they have? Now, I think
the answer is this is "Yes". That is why I make
the further observation that such things as this may
have a great effect on this whole matter of classi-
fication. There are different bases and these have
to be recognized before we can get this thing on an
international scope.

Mr. Green: I assume you would agree, Mr. Knight?

Mr. Knight: Yes, Mr. Moderator, I think our op-
position procedure has a great deal to do with this,
because in the opposition proceedings the opponent
may bring in a specification from any other country
or any other piece of literature he likes. That is,
in effect, part of the official proceedings before the
grant of a patent. It can be regarded, in a way,
as part of the official search, although done by the
opponent and not by the office.

Mr. Purdy: It seems as though we keep coming
back to this divergence of interests in the search.
Mr. Rubach and Mr. Knight both mentioned that
within their classification systems there are two gen-
erally recognized kinds of classes, the general class
and the class which is based more on art or tech-
nology and, as Mr. Knight put it, the general and
the application class. Whereas Mr. Vincent pointed

111

out that our type of classification is tending more and
more to be based upon proximate function which
corresponds to Mr. Rubach's construction, function
or purpose group, I believe. Now, there is no rea-
son, of course, why we can not have two complete
schemes of classification in which every docu-
ment could be filed, except of course, for the over-
riding one of the additional space involved. And
I wonder if it is possible, or if it has been attempted
any place, to classify documents for hand searching
on one scheme and follow it up with a machine
search in the more naive scheme. I wonder if this
gentleman who so kindly defended the business
machines has had any experience in that direction.

Mr. Fairthorne: I do not know of any such thing
done in that way, but the general feeling you have
expressed, that a machine could do the rough sort-
ing and human beings then get down to details, is
one that is very commonly felt by even the most
ardent machine defenders on the ground that you
can produce a high concentration of what you want
in a short time, and then get down to look for de-
tails. It is what Mr. Newman has said. You first
look for roughly similar and then you look for the
differences. That sort of thing is certainly possible.
But all these involve impractical application, par-
ticularly in the Patent Office work, or any National
Library. Enormous storage of group data has to
be rapidly accessible. And rapidly accessible does
not mean by writing it on tape, because writing on
a piece of string is remarkably silly, even if you do
write it with a magnet and pull the string very fast.
So that we are waiting to have a great deal of much
more rapid access static store in machines. I myself,
this is a highly controversial point, feel that we can-
not do that sort of thing really well on a big collec-
tion until we have some ten to the ninth valued
digits in fairly immediate accessible form. That is
an American billion, not a European billion.

I think we can do what you think, on a rational
size collection. If you take a smaller collection, it
is always quicker to do it by hand. That is one of
these troubles. The small scale thing is not a proper
model of the big one. And that is why we are hav-
ing great difficulty with the mechanical search. We
are too big for the small collections and too small
for the big ones.

Mr. Pfeffer: In the Mechanized Examining Divi-
sion of the Patent Office this goes on to some extent,
particularly in the art dealing with steroids. Some-
times the examiner cannot ask a question which

would give him a complete answer on the mech-
anized system, and he will very generally supple-
ment his search by the use of the manual
classification system. This is not unknown. Of
course, this can de done with a system as small as
that,' but I have my doubts with respect to a very
large scale system.

Mr. Purdy: I think that Mr. Pfeffer is giving us an
example of what is more or less the reverse, or in-
verse, of what I was thinking of. You are probably
examining or searching the more complex items of
information with your machine, and going to the
existing classification for use perhaps.

Mr. Pfeffer: No, I would say it is the other way
around. The system is rather simple in its concept,
and the more complex things have to be searched
by hand, and the simpler things can be searched
by machine. We have not yet reached the stage that
you are proposing.

Mr. Green: I wonder if we have a problem yet that
we can work on cooperatively. And this it seems
to me is the theme of this afternoon's discussion.

Dr. de Haan: Mr. Moderator, I should like to have
some answers on whether I am right in saying that
our existing hand classification systems are so heavily
loaded with tradition, that it seems very difficult to
come to a general system for hand searching. But
it may be possible that in the field of automation,
since we start from the bottom anew, that we could
come to an intimate collaboration in order that we
get there altogether in the same line, developing
this automatic search system which still has to be
created.

Mr. Green: This seems to me vital. Are there any
dissents? Do you want to make an observation?

Mr. Urbach: Not really a dissent. That is almost
like condemning motherhood. I would say that I
can see a problem coming in that area in line with
some of the present research being conducted by the
U.S. Patent Office group, in that there is a tendency
to automate within the units of the existing classifi-
cation structure, as opposed to looking at the entire
area and automating all of it. If this tendency ex-
ists also in other patent offices, and I would guess
that it does, we do not solve the problem by going
to automation. We just substitute another tech-
nique for the existing techniques within the existing
classification systems that are not compatible with
one another.

112

Mr. Green: This is certainly true, but I believe Dr.
de Haan is saying, that now is the opportunity to do
something about this, and I think ( I am not a mem-
ber of the Patent Office, but I am confident that I
am speaking for the Commissioner when I say) that
the U.S. Patent Office would be happy to cooperate
in an international group on this sort of thing.

Mr. Pfeffer: I think if we are going to deal with the
problem of research, which Dr. de Haan must con-
cede is the important thing here, that it would be
rather difficult to try to allocate areas of research
and say that, you work on this problem, and you
work on that problem, and I'll work on the third
problem. I think probably the best way to ap-
proach this would be on a team effort, where each
of the contributing offices would supply manpower
and have a unit, an international unit if you will,
which will consider this problem, and be supported
by each of the contributing agencies here.

Mr. Green: May I also suggest another device, the
guest worker device.

Dr. de Haan: I think that the solution of automation
is not one problem. I think that perhaps in chem-
istry, in electronics, in the mechanical field, you may
ultimately have to use quite different systems, and
perhaps the study of various problems which the
experts must divide and must arrange, could be
divided amongst various patent offices. They must
give all their reports of their studies. They must
interchange. In this way perhaps something might
be done. I quite agree that an international com-
mittee would be very good, but the Department of
Finance will intervene. At least in the European
countries.

Mr. Green: Here, too. Our limitation this after-
noon is problems in patent classification, but I as-
sume that after the third day there would be some
rather specific plan for collaboration in this area.

Mr. Lawlor: I would like to ask two fundamental
questions here. One is a mere repetition of ques-
tions that have already been asked, and that is:
Why can't you have multiple classification schemes
applied to the same document? I see nothing
wrong with this. It is a good thing. Then I
would like to ask another fundamental question, and
that is this: Is classification necessary? I am an
advocate for assigning as many problems as pos-
sible to machines. I am an advocate of the full
text recording system. I am an advocate of the
system in which you would record the full text of

the patent, or the full text of the key words that you
assign to the patent, or the full text of the sentences
that you use in abstracting the patents, and you
never classify anything until you ask the question.
When you ask the question, you ask the question in
terms of the language that has been recorded. You
find all of the pertinent documents and your search
results are not prejudiced because somebody made
an error in classification. Now that is really a
question, not just an exposition. Is classification
necessary?

Mr. Green: Shall we give our foreign guests a
chance to answer first? Are there any in the group
from abroad who would like to comment on this?

Mr. Fairthorne: This is certainly a clean problem in
a sense, but the trouble is that it will hurt for prac-
tical as well as theoretical means of solving it for
some time. But there is no reason why one should
not start trying to do it now. I think anybody who
thinks it is going to be a help in the next decade is
perhaps rather optimistic. On the other hand, it is
in the long run the only way you can deal with any
retrieval problems at all, because the only thing
that is constant in time is the marks you put on the
paper. We are dealing with the marks that people
make in a claim, in a book, or something of that sort.
They are a set of marks. What we want to do is to
tell a person or a machine, at a given time, how to
recognize when he looks at the marked object that
it is what he wants. We want to delegate to the
machine the fact that we can recognize that the
patent or document is relevant to whatever we have
in mind. And that is very difficult. We do not
know how we recognize it. But it has one great
advantage. It does not involve prediction of the
future in any way whatsoever. At the moment,
not being able to do that, we do not know how to
recognize things nor have we any means of looking
at complete texts in a hurry. We get somebody to
write on top of the document, marks which indicate
how he hopes it will interest people in the future.
Now there is not a fixed amount of information in
a document, that you can squeeze out of it, like
water in a sponge, although the electrical engineers
believe that. What electrical engineers believe
about these things is a matter we are not discussing
here. There is just as much in a document as
people are interested in at a given time, and, there-
fore, we cannot avoid having a good deal of pre-
diction in any indexing or classification, which is a
past expression of hope. And we can not avoid

113

leaving things out because the things we are really
classifying are not marks on a document, but how
we think people may be interested in it, and that is
indefinitely large. So that, when we aim at treat-
ing the original text to find what is in it, we are not
only following what we do ourselves, (most of us in
fact read the text before we find out what it is about,
and we are so used to it, that we do not think about
this), but we are cutting out an insoluble problem,
logically. Unfortunately, we will still continue, I
am afraid, to have to make marks in a systematic
fashion on top of documents, to indicate how we
think they are going to interest people in the future,
for quite a time. But do not let that blind us to the
fact that, as we pointed out, in the long run we must
be able to delegate to unintelligent agents, tell them
how to recognize what we want in a printed text,
because that is what we ourselves do. The text is
there and it is the only thing that is permanent
over the ages, barring wars and fadings. But I
think all the work that has been done on this is
very worthwhile. It will have to come some time.
One big problem is mechanical handling of your
texts, which come in all shapes and sizes and that, as
a mechanical job, is almost as bad as the intellectual
one. But there we are.

Mr. Green: It seems to me that those comments
were summarized for me by one of the research
examiners in the Patent Office when he said that the
key word here is "riot". I said, "What do you
mean by 'riot' "? And he said, "Retrieval in our
time."

Mr. Purdy: May I try to answer the easier of Mr.
Lawlor's questions? The reason why, I think, most
Patent Offices, and I am sure it applies to our own,
do not have dual classification is the space reason.
If we had to add an additional document for two
out of three we should have to make room for two
million patents, and we are already bulging at the
seams in the research room. We have tried to
create dual classifications on a very highly selective
and I should say guarded basis, which will help the
searcher who is not interested in the same field that
the patent examiner is, that is, duplicating some-
thing set before him, but who just wants to know
what the state of the art is.

Mr. Green: Any additional observations?

Dr. de Haan: Mr. Moderator, I might ask of the
speaker before the last one, is it not true that the
mechanical search system makes it possible, through
the machine, to have an unlimited number of

approaches to the material contained in the
documents?

Mr. Green: Do you care to speak to that?

Dr. de Haan: I mean that if you now have the
hand classification, you have one approach, either
through function or technology. But is it not true
that in a mechanical system you have an unlimited
possibility of approaches to get something out of the
patent?

Mr. Green: You can enter the system in a number
of ways.

Mr. Fairthorne: I do not think there is any doubt
that given two conditions, that we can do this sort
of thing up to a point by machine, far better than
we can by using highly trained people. The con-
ditions are first of all, of course, that the stuff can
be read by a machine in some way or other. By
stuff, I mean probably the index marks of people
that have been made. But given a rational method,
a reasonable method of marking up a patent to
indicate our current interest in the world of today,
(that is by any of the standards of classifications
used by the various countries ) you should be able to
approach these marks from different points of view.
Although there might be a great deal of computing
to be done, real computing, in the way of changing
structure, words and things in a way in which the
linguists and others are trying to analyze now. We
are, in fact, trying to see how people use words.
Now, I am not a linguist. I am a mathematician.
But I think even on simple matters, there is no
doubt, that given the rules of recognizing the marks
are what you want, and the fact that they are what
you want, (which is not always clear) a machine of
the type we have today can deal with this matter,
although approaching it from different points of
view much more quickly than a human being, in
fact, does so. It is not for me to advertise my
particular affiliation at the moment, but we do it
pretty well. But mind you, it means a great deal of
intellectual effort to put the marks on. But once
they are on, we can get it out not as well as we
would like, but we can do it.

Dr. de Haan: I think we cannot avoid putting in
human intelligence somewhere.

Mr. Fairthorne: We do not want to use the same
intellect on the same job over and over again. That
is how the people have mapped it out. At the
moment we have to, I would say, but the idea of
having a common classification office, common inter-

114

classificating thing is to avoid intelligent, knowledge-
able people having to do the same work over again.
That is why we are here.

Dr. de Haan: Exactly. I think even with mechani-
cal searching you have to analyze the documents.

Mr. Fairthorne: For a very long time. Current ex-
periments are rather naive, and sometimes, if I may
say, purely silly.

Mr. Green: It seemed to me for a long time that
this problem of preparing the patent for machine
storage is ignored to a great degree by the people
who want to sell you a system. They say, "Once
you have prepared the material, we will give you
the system." But the main problem is getting the
material together. And this should not be over-
looked.

Mr. Urbach: In connection with Dr. de Haan's
point again, of unified research effort, I would like
to ask if there has been an exchange of what research
efforts already have been made by the various pat-
ent offices, and if so, how does, let us say, a partic-
ular mechanized American system work in a
particular foreign patent office, and if this has not
been done, is this not a good place to start?

Mr. Green: My understanding was that it had not
been done, but this is the time to start.

Dr. de Haan: But we have done so already. We
have developed in our office, a mechanical system for
searching carburetors and we have prepared a deck
of cards. Mr. Andrews generously offered to try
them. He now has the cards and he will complete
the test in a short time I hope. But we have already
made a study on this, just in the mechanical field
because we thought that it was much more difficult
than the chemical field, and we wanted to see if
there was a possibility. We are very anxious to
know the result of our work, because we have not
at our disposition a machine which is sufficient for
handling the cards we have made.

Mr. Green: I should mention that every representa-
tive of a patent office who attended this Conference

was sent a questionnaire in advance and he was
asked a number of questions, one of which was
"What mechanization efforts are going on?" And
this material came in, as you can imagine, at differ-
ent time intervals and was just assembled last week
into a very useful summary which, regrettably, is
not yet available. But I have seen it, and it will
be in the Proceedings and it gives the sort of infor-
mation you are asking about.

Mr. Urbach: I do not think I have quite made my
point. Perhaps I think there is more mechaniza-
tion today than there really is. And that is the
answer to my point. But I know that there are in
use today, a number of small systems in the U.S.
Patent Office. I believe that is the case with some
of the foreign patent offices. Have there been, or
should there not be, exchanges of these systems.
Perhaps we are going to run into problems where
a system designed for one patent operation will not
work well in another, as might be the case with the
concordance. Shouldn't we find this out now, by
exchanging total systems, which may be nothing
more than a matter of transferring punched cards?

Mr. Green: I think this is a useful thing to consider.
Again, I would call to your attention that we are
not talking about complete systems. We are talking
about problems of patent classification this after-
noon. I had hoped that we would limit our discus-
sion to that. Of course, at the end of the three days
something of this nature, I would imagine, would
come forth.

Mr. Pfeffer: May I call Mr. Urbach's attention to
the program for the morning of the third day,
Wednesday, in which this information will be
brought forward.

Mr. Green: Possibly we are getting to the point of
diminishing returns. Unless there is any objection,
shall we adjourn? I might say that you are to come
together tomorrow morning here at nine o'clock. I
have enjoyed the opportunity. Thank you.

The second session was adjourned at 5:30 p.m.

115

Third Session

Tuesday Morning, October 24, 1961

The third session was called to order at 9
o'clock by Mr. Harold Pfeffer, Office of
Research and Development, U.S. Patent
Office, Moderator.

Mr. Pfeffer: Our first speaker this morning is Dr.
de Haan, the President of The Netherlands Patent
Office. Dr. de Haan received his legal degree in
1924 from the University of Amsterdam. He grad-
uated as an electrical engineer in 1930 from the
Technical University of Delft. He became a patent
lawyer, practicing at The Hague from 1932 until
1947. In 1947 he was appointed President of The
Netherlands Patent Board and Director of the Neth-
erlands Office for Industrial Property, and has
served in that capacity ever since. He received an
honorary Doctor of Law degree in 1946 from the
University of Amsterdam. I am pleased to present
Dr. de Haan, who will speak on a suggested interim
solution to the patent examination problem. Dr.
de Haan:

Dr. de Haan: Thank you very much, Mr. Pfeffer.
I will make one observation here. You mentioned
that I received an honorary degree in law. That,
I am afraid is not so; I had to work for it.

Ladies and Gentlemen, I am very greatly honored
to be invited to speak here about the modification
of the Netherlands law, which is at present before
our Parliament. I hope you will excuse me for hav-
ing to improvise this speech as I learned only last
week that I was to speak here. Due to the great
hospitality of our American colleagues, I did not
have the time to prepare a paper from which I could
read. Besides, I am obliged to speak in a language
that is not my own, which is always a bit difficult.
I hope I shall come through. When I am not suf-
ficiently clear, please be so kind as to interrupt me
and ask me what I mean to say.

Coming now to our new proposal that tends to
certain modifications of our present system, I think
I can be most clear by explaining our present sys-
tem of patent law first. I will go back somewhat
in history and start by saying that our patent law

was accepted in 1910 after very long preparation.
This lasted, if I am right, from 1896 until 1900,
when it passed Parliament. As a remarkable fact
I should like to mention that this Netherlands pat-
ent law was drafted for the greater part by a highly
respected committee of three people : a famous law-
yer, a famous civil servant and an economist. These
three/people had no experience at all in the field of
patents. They were very intelligent, highly intelli-
gent as a matter of fact, but they never visited for-
eign patent offices existing at that time. They were
guided by the existing patent laws of Germany,
Great Britain and the U.S. They followed mostly
the example of the German patent law. The fact
that these men succeeded in drafting a law in such a
form that until now it has completely satisfied
everybody, is really remarkable, I think. This law
was accepted by Parliament in 1910 after which the
Patent Office was established. It started its activi-
ties in 1912, at the same time that the law entered
into force.

During the time between the two wars the Patent
Office also had the good fortune of having a Presi-
dent of great ability, a man who was a wonderful
organizer and who built a machine so well that it
could withstand even all the troubles of war. Mr.
Alingh Prins, perhaps some of you remember him —
he was the President of the Hague Conference of
1925 — is still alive and is now 87. As you see, the
patent business is a healthy business. During his
time in the Patent Office Mr. Prins was fortunate
enough to find the economic situation between the
two World Wars such that we could get very good
engineers. So before the last war the Patent Office
was able to have as many young engineers as it liked
and it could pick and choose among them. Thus
the Patent Board started out with the most outstand-
ing technicians who, much to my regret have now

116

reached an age where they will have to resign within
the next years. One of the most outstanding of
these brilliant technicians is my friend, Mr. Bulder,
who is here amongst us. So, due to good law and
a good staff, the Patent Office was able to handle
the new practice of granting patents, according to
the system of the law.

What is this system? It is the following: it is
divided into two procedures. We have, first of all,
the handling of patent applications by the Bureau of
Industrial Property. And afterwards the handling
of the application by the so-called Patent Board.
These are two formally divided institutions. I my-
self, have the honor to be Director of the Bureau of
Industrial Property; as well as being President of
the Patent Board.

The Institute is solely an administrative body,
whose duty it is to make the novelty search and the
preliminary search of patent applications. The in-
coming application is handed to an examiner who
looks to see whether it complies with the stipulations
of the law. As you all know, the most important
of these investigations is the novelty search. In
general, the patent attorneys know the legal stipula-
tions of the law, so usually they will not file an ap-
plication which is against these stipulations. But,
to our great regret, they do not know and cannot be
acquainted with the prior art, so that we have to
investigate what there is in the way of prior art.

The examiner of the Netherlands Patent Office
works like his colleagues in other countries. He
states the prior art and informs the applicant ac-
cordingly. The applicant has the right to refute
his arguments. The applicant can file new claims,
and only when the examiner is of the opinion that
he has sufficiently searched the application, he sends
it to the Patent Board, which consists of thirty mem-
bers, of whom five are lawyers and the others,
technicians.

The member of the Patent Board who receives
the application studies the case. He considers
whether he can agree with the advice of the exam-
iner, and if that advice tends to a decision to lay
the application open to public inspection, and the
member of the Patent Board is in agreement with
this, he informs the applicant accordingly. In many
cases this will be done only after the applicant ful-
fills minor changes which the Board deems neces-
sary, after which, without further procedure, if the
applicant agrees, the application is laid open to
public inspection.

But, if the Board is of the opinion that the appli-
cation should not be laid open to public inspection,
because the Board is of the opinion that the appli-
cation lacks novelty or invention, then the applicant
has the right to defend his application before the
Board. If the defense of the applicant before the
Board, which in this stage consists of only one mem-
ber, does not succeed, and the decision is in the neg-
ative, he has the right to appeal to the Board of
Appeal, which consists of three members, two
technicians and one lawyer. After decision is made
by the Board of Appeal the application is either
laid open or definitely refused. Against the decision
of the Board of Appeal, tending to non-granting of
the patent, no appeal is possible. It is final.

After the application is laid open to public in-
spection, anybody can launch an opposition proce-
dure during a period of four months time. This
opposition procedure comes, in the first instance,
before the Application Board, which in that case
consists of three members, the member who has first
decided to lay the application open to public in-
spection, another technical member, and a lawyer.
After a decision is made the party who has not
succeeded, has the right to launch an appeal which
comes before a Board of Appeal also consisting of
three members; in some complicated cases a Board
consisting of five members may be nominated, the
findings of which are the final decision of the
Patent Office.

As you see, there are cases in 'which the same
application may come before the Board four times,
twice before the Board of Application and twice
before the Board of Appeal. I think that this is an
excellent procedure which guarantees the applicant,
and third parties, that the applications are treated
with the utmost care.

You will understand though that such a proce-
dure requires much manpower, and we all know
that at the present time manpower is very scarce.
I am quite confident that the Netherlands Patent
Office is not the only office in the world which
suffers from a tremendous lack of staff. And
through this lack of staff, we have experienced a
tremendous arrear in our work for the past ten
years. I believe there are at present about 40,000
applications in the Patent Office not yet disposed
of, and the number of incoming applications is at
the present moment about 14,000 annually, so on
the average, we have an arrear of three years. This
includes applications which are to follow the com-
plicated procedure I outlined for you and which

117

may be in arrear of ten years from their priority
date.

About five or six years ago, I advised the Minis-
ter that we could not continue this way, although
at that time the arrears were much less than nowa-
days.

Our Minister appointed a Commission in which
not only the civil servants, but also the patent law-
yers, patent attorneys and representatives of the
industry were represented. We have worked very
hard in this Commission to find a solution. As you
must have found yourselves, modification of law is
not easy and it consumes a lot of time. Finally,
after many modifications, we arrived at a project
which has now been submitted to Parliament and
which, as far as I know, has met with unanimous
approval.

What is the underlying idea of this proposal?
First of all, I should point out that the interested
circles in The Netherlands were very keen on main-
taining the system of the present law as I sketched
it to you some moments ago. Although the system
is very complicated and requires much labor, in-
terested circles are happy with it. At first we
thought that this might hamper any modification,
but the work had to be done. After due reflection
we found statistically that the average life of a
granted patent is about five years. As you know in
The Netherlands we have renewal fees, and though
I do not wish to bring into the discussion the ad-
vantages or the disadvantages of renewal fees, one
thing is undeniable, viz, that the obligation of pay-
ing renewal fees serves more or less the purpose of
cleaning of the registers. I mean to say that if
someone is no longer interested in his patent and
he does not pay the renewal fees, the patent is au-
tomatically dropped. Consequently the average
life of a patent is an indication of the interest which
the applicant will have in his eventual patent.
Those of you who attended the panels last week ob-
served that one of the gentlemen present here spoke
about the same question, saying that it is rather
seriously doubted that all the work done by Patent
Offices is justified since so many patents are of no
economic value to their owners. He also mentioned
the fact that in the United States the idea had come
up to establish a comittee to judge the interest —
the economic interest — of inventions first of all. I
quite agree with the speaker of that day when he
said that such a committee could not work. It is
quite impossible to say before hand whether an in-
vention will really have economic interest. All the

same, if an inventor, who has made an invention
which has, perhaps, not yet been worked out in all
its details, but is a workable invention (as I said
one that has utility and that is new) , which he wishes
to be patented, but he does not know whether it
will be of any interest to his industry, he is obliged
to file an application, especially in European coun-
tries where the priority date is not the date of the in-
vention, but the date of application. In European
countries patents are granted to the first applicants
and not to the true and first inventors. So he must
put in an application and only later will he be able
to see whether the patent granted on his applica-
tion is of any use to him.

Taking into account the fact that the average life
of the patent is five years, we considered whether it
would be possible for the applicants themselves to
decide whether they wanted to have a patent im-
mediately, or that they should wait with requesting
the granting of a patent until they had more cer-
tainty about the economic value of their patent.
This idea is the basis of our Bill. As you see, this
Bill does not change anything fundamental in the
existing procedure of the Patent Office; the change
being merely that the applicant himself can decide
whether the procedure, first the preliminary search
and in the second instance the granting procedure,
will be started. By this method we hope that our
restricted manpower can be devoted solely to those
applications which are of real interest, while those
applications of which interest for the applicants is
not certain can remain dormant until the moment
interest becomes apparent or until they become
dead files.

We have the following stipulations in our pro-
posed Bill : If an application has been filed and the
applicant requires nothing of us, we do nothing and
the application remains dormant until the moment
that the applicant asks us to proceed with the
novelty search in order to make him acquainted
with the prior art. After we have supplied him
with the prior art, the application rests again and
we hope that because of the prior art, the applicant
will abandon his application. We do not hope so
for the applicant, but for the furthering of our work.
But if he thinks that the prior art we have supplied
is such that there is a real chance for a patent, he
will ask us for the granting of the patent. If seven
years after the application date nobody has asked
for either the novelty search or for the granting pro-
cedure, the application will be dropped automati-
cally. We have tried to come to terms on a period

118

ml years, somewhat longer than the average life of a
patent in The Netherlands, hoping that in this way
i number of non-important inventions — non-
important from the economical point of view will
iie out. We have to keep in mind, as you have
remarked yourselves that some technical, very simple
inventions may turn out to be great successes, while
Dtherwise technically remarkable inventions may be
:>f no economic value.

The interested circles, especially in industry had
3ne more objection, against the situation that was
created by our extensive arrears. They said, "It is
unbearable that we sometimes have to wait three,
four or five years before the pending applications are
ready to be laid open to public inspection. We do
lot know what we should do if we want to start on
a certain manufacture; we do not know whether the
Patent Office will grant a patent on another pend-
ing application which dated from the time before
we started our manufacture, and which obliges us
to stop our manufacture as otherwise it would in-
fringe the afterwards granted patent."

So, we decided that all applications will be pub-
lished eighteen months after their priority date.
This way the industry is in a position to know which
applications are pending, independent of the fact
whether or not the search and the granting pro-
cedure have been requested by the applicant. As
you will understand, if industry notices that an ap-
plication has been filed and published, and possi-
bility of conflict with a future patent is present and
the applicant happens not to ask for the granting
procedure, industry has a real interest in certainty
and on starting the search and granting procedure
on its own behalf. For that reason we have intro-
duced into our project the possibility also for third
parties to start the granting procedure as well as
the novelty search. We do think that this way we
have not interfered in any way with either the in-
terests of the applicants or with the interests of third
parties, while at the same time we have complied
with the wish of interested circles not to alter the
fundamentals of our present system of patent
granting.

Mr. Moderator, I have mentioned the most im-
aortant facets of our system of modification. If I
have failed in any way my friend Bulder will correct
me. Thank you.

Mr. Pfeffer: Thank you very much, Dr. de Haan.
ff we are to judge from Dr. de Haan's presentation,
I would say that the hospitality on the part of the

host is a very fine substitute for work on the part of
the guest. Thank you again for a fine presentation.
Now, are there any questions of Dr. de Haan?

Mr. Michel: Dr. de Haan, after the search and the
granting procedure have been gone through, and
you grant the patent, does the life of the patent date
as it did before, from the date of filing?

Dr. de Haan: Mr. Michel, it did not date from the
date of filing before. It dated eighteen years from
the granting date. I forgot to say that there was
also the objection of Industry that through the early
years of a patent grant it usually became effective
ten years after the filing date, and then still existed
eighteen years from the granting date. So in our
new system we have brought in the modification
that the patent lives twenty years from the applica-
tion date, or eighteen years from the granting date,
which ever is most advantageous for the applicant.

Mr. Michel: Upon request by the applicant or by a
third party for search and granting procedures, are
there any additional fees required?

Dr. de Haan: There are. But I think that the fees
required by the Patent Office will not be of great
importance, and are quite justified, because the fees
that the applicant has to pay to his patent attorney
are much higher than the official fees of the Patent
Office. You know, that is not a reproach against
the patent attorneys, because they sometimes have
to do quite a lot of work.

Mr. Michel: Where a third party is asking for the
search, who pays the fees?

Dr. de Haan: The third party pays the starting fee,
for starting the procedure.

Mr. Helgeland: If this system should work, let us
say seven years, you might then have patent applica-
tions lying there, resting, for which nobody has
asked for a search. You then get an application
and you ask for a search. Do you tend, then, to
search among the 40,000 resting patent applica-
tions, to determine whether there is an older filing
date or older application colliding (Interfering,
Ed.) with the application for which a granting pro-
cedures is now asked? And, I suppose that all ap-
plications, whether they ask for granting procedure
or not are to be classified and divided among the
examiners, and are lying there and every seven
years, the examiner takes away the lowest. Shall
the examiner now, if he gets an application asking

645803 0-63-4

119

for granting procedure search in this file to see
whether there is an older patent application collid-
ing more or less with it, and who shall pay? If he
finds for instance ten such applications shall he start
examination of these ten applications and who shall
pay for that?

Dr. de Haan: That is really a difficult question.
But I think we can start the examination because
our whole procedure requires that we must have
laid the application open to public inspection and
then it is up to the man who has an older right to
make opposition. And do not forget, we had the
publication of the application, so the examiner can,
under the older applications, look for colliding ap-
plications and inform the applicant, during the
search, that there is a colliding application.

Mr. Helgeland: But if the owner of the older ap-
plication does not do anything about that, does he
lose his right?

Dr. de Haan: No. Because the older man whose
application is published is the first applicant. So
as far as the second application collides with the
first application, he has not the right to a patent.
So our searcher must do' a search among the prior
art until the filing date, and among the published
Netherlands applications, also until the filing date.
He can state that there is a prior application.

Mr. Helgeland: But does the examiner have an
obligation to search among his files of resting
applications?

Mr. Bulder: But they are published. They are in
the pigeonholes. All published material is taken
into consideration when we search for novelty.
But, nevertheless, we are obliged to search for it and
to cite it.

Dr. de Haan: As far as they are not published, cer-
tainly he must look in his files. But we already
have in our Classification Division a certain system
to trace colliding applications. I will not say that
it works perfectly, but we can generally tell, al-
though it is now much more difficult with our ar-
rears which are not published to trace colliding
applications than it would be under the new system.

Mr. Michel: In that case, the onus is not exactly on
the searcher or on the Patent Office if during the
search on an application the searcher or the office,
finds that there is present in the office an applica-
tion for the same invention having been filed prior

to the one he is searching. He can refuse the later
one on that prior one provided it has been published.

Mr. Bulder: Yes, certainly he can. And if he waits
six months it is always published.

Mr. Jeffery: When you say that the specification is
laid open to public inspection, does that mean that,
in law, it is said to be published. We have in Aus-
trailia a procedure of saying that a thing is open to
public inspection by just publishing the number in
the Journal. Is it like that, or it is a publication,
a printing?

Dr. de Haan: It is not in the law, but it will be in
the rules. That is a difficulty. We now have ini
this system two kinds of publishing. First of all,
eighteen months after priority date we publish it as
it was filed, and if the granting procedure has been
asked, we publish it again in the way in which we>
are prepared to grant the patent.

Mr. Jeffery:. Since you have made a first publica-
tion, does not this put a restriction on the right of
the applicant to amend and alter it?

Dr. de Haan: Formally not. Practically perhaps.
Because it very often happens that we state after-
wards that the applicant has succeeded in introduc-
ing new subject matter into the application, which
under the law is forbidden. If we have made sucht
a mistake we now have the help of opponents who
will tell us that such and such a thing was not in the
original application. So it helps us to comply more
with the law than perhaps we do at the present
moment. I will not say that it is easier, but it may
happen that it can now be controlled by everybody.

Mr. Michel: In the case of an application being
filed in the Office originally, naturally, as I under-
stand it, it would have to have claims. What
happens if the applicant claims several inventions in
that application?

Dr. de Haan: At the moment of filing nothing hap-
pens. But at the moment of searching, when it is
not before the Patent Board, but in the Office, we
can say that more than one invention is being
claimed and that the application will be divided.
This has been the subject of a lengthy discussion in
our Commission because, not without reasons, the
patent attorneys argued that the Administration and
not the Board should have the right to make the
decision. We have found the following solution.
The Administration, the Institute of Industrial

120

Property can say that there are several inventions
Jn the application, and they have to file a second
application. But, if later on when the granting
orocedure is started and the Board thinks that it
vas an unjust decision, he gets back his fees. On
hese terms we have come to an agreement with the
nterested circles.

Mr. Michel: Where there are several inventions
:laimed under one application, and a third party
decides that he wants a search made to find out
whether his manufacturing of one of such inven-
ions might infringe a subsequent patent, has the
ipplicant the onus to ask for a search and abide by
he divisional requirements of the Office and carry
mi the examination on all the inventions that he
las claimed?

Dr. de Haan: The third party pays the fees for the
search.

Mr. Michel: Yes, but he may be interested in only
>ne invention being claimed there, so he would pay
he fee for that part for which he is interested?

Dr. de Haan: Yes.

Mr. Michel: But the rest is not being searched?

Dr. de Haan: No.

Mr. Helgeland: What do you do with the applica-
tions for which granting procedure has not been
requested? Are they tested only for the formali-
ties? Are they treated almost like the treatment of
patent applications in registering countries?

Dr. de Haan: Yes. The only thing we look for is
that they comply with the formal stipulations of law
and rules: that there is a specification, that there
are claims, that the filing fee is paid, that there are
so many samples of specifications, and claims, and
so on. But if an application, and this is also true
in our existing system, does not comply with the
formal prescriptions of the law, we do not consider
it an application at all. That is only a sheet of
paper.

Mr. Helgeland: Isn't it possible then that the system
would throw the gates wide open for misuse, in that
vague and broad applications would be filed and
probably, as I understand, for a minor fee only?
Isn't there a danger that industry, the applicants,
being so minded, could fill up your file of resting
applications with vague and obscure patent applica-
tions, in order that they may have the possibility of
picking out of their file anything to object against
if anything comes up during seven years?

Dr. de Haan: We have considered this matter.
You understand that if you have a lot of obscure
applications filed, which may be possible, it is espe-
cially in the interest of the Industry to see that it
does not happen. So we had in our original project
a stipulation, that if such un-understandable appli-
cations were to be filed, the Patent Office would
have the right to initiate the searching, and have
the fees paid by the applicant, for the reason that
the application is un-understandable. But the In-
dustry was against such a stipulation and it is espe-
cially because of the interest of Industry that it is
not there. We complied and dropped the stipula-
tion. So I think the danger is not so great if the
interested circles do not feel it.

Mr. Helgeland: It is just the interest of the Industry
that I might fear.

Dr. de Haan: The applicants might have an interest,
but the serious industries have no interest at all in
filing such applications and for that reason, third
parties can always ask for the granting procedure.

Mr. Pfeffer: Dr. de Haan, may I ask what sort of
general reaction have you had from the public on
this procedure?

Dr. de Haan: Well, we discussed this whole matter
for a long time with all the interested circles we
could trace, and until now it seems to me that we
have had, in all, general approval. So we hope
that it will pass through Parliament without much
opposition. But you can never predict what will
happen in Parliament.

Mr. Michel: With this system, did I understand you
to say that the average life of a patent was about five
years with renewal fees? What is the percentage
of time devoted to search that you hope to save? Do
you expect to make only about half the searches that
you were making before, or less or more? Of
course, you have not tried it. This is just a guess.

Dr. de Haan: I must say that it is absolutely uncer-
tain if this system will be a success. There are peo-
ple who say, "You will save only five per cent" and
there are people who say "You will save fifty per-
cent." It depends completely on the applicants. If
all applicants are keen to get their patent immedi-
ately, the system will be of no use at all. But as far
as our greater Industry goes, they say they will cer-
tainly make use of this system to a great extent, be-
cause they really have many applications that they
do not know what to do with at the moment of nlin°;.

121

Mr. Michel: Yes, there are a great many applica-
tions filed in Europe I notice, and probably in every
one of our patent offices, as safety measures. They
do not know if they are even sure that they will ever
use it, but they file an application just to make sure
that they come in first to prevent others getting a
patent, blocking the field.

Mr. Rot kin: I understood you to say that the appli-
cations are handled by various examiners on a clas-
sification basis. Are they assigned to particular
experts or specialists in the area?

Dr. de Haan: Yes.

Mr. Rotkin: To what extent, in your Office, are the
examiners free to exchange their workload. By
that I mean, that if after initially examining an
application he determines that someone else is more
expert in that area, can he freely transfer the appli-
cation.

Dr. de Haan: The direction of the application is a
form. It goes to the Classification Division. The
Classification Division writes on — it is not exactly
so, but I would go into too much detail to say it
otherwise — the application a Division. We have
thirty divisions. And then the examiners in the
Division decide who will take the application. Or
they may say it is not for our Division. It should go
to another Division. In all these classification
questions, the Chief of Classification, who is Mr.
Koelwijn has the final decision. But it is very, very
exceptional that the examiners do not come to an

agreement among themselves as to who will take th
application.

Mr. Pfeffer: I realize there is a good deal of curi
osity about the different procedures, but I think fc
the sake of the time element, we should confine oui
selves to the particular question at hand.

Mr. Reznek: Who are these third parties? Ca
they be the assignee of the application? That is, i
a man has an interest in an application and wanl
the patent to be issued, can he act as a third party
Can he pay the fees?

Dr. de Haan: Anybody can act as a third party. 1
is open to everyone to ask for the granting procedure

Mr. Kalikow: Can you say whether or not there i
any serious consideration for using the Netherland
system of such plan as the basis for Common Marke
patents?

Dr. de Haan: I think that it still has to be workei
out, but I think the idea of the Common Market i
that you have automatic novelty search and the dt
layed granting procedure. In our propositior
project of law, there is the choice for both instances
but I think in the Common Market they are mor
inclined to have an automatic search and a delayd
granting.

Mr. Pfeffer: Any other questions?

Thank you, again Dr. de Haan, for an interestin
morning. I would like to suggest that we adjoun
proceedings now for a coffee break.

The third session adjourned at 10: 30 a.m.

122

Fourth Session

Iuesday Morning, October 24, 1961

The fourth session was called to order at
10:45 A.M. with Dr. Watson Davis, Direc-
tor, Science Service, Moderator.

Mr. Pfeffer: We will begin our fourth session this
norning on Problems in Abstracting and Analysis,
ind your Moderator for this morning is Dr. Watson
Davis. Dr. Davis is very well known amongst most
)f us here and needs very little introduction. I
hink most of you will know him as the Editor and
Director of Science Service, with which he has been
:onnected since 1921. Dr. Davis is a civil engineer
)y profession, and has an honorary doctor of science
legree from George Washington University. He
5 also a Past President of the American Docu-
nentation Institute, and he has done much to help
levelop the use of microfilm as a means of distribut-
ng scientific information. I would like to present
o you Dr. Watson Davis.

)r. Davis: Doing this job gives me great pleasure,
specially since I can turn to Mr. Pfeffer and ask
dm what I should do. It seems to me that what
r ou people are discussing here this morning is of far
[reater importance than these 9 million patents that
lave been granted, and the influx of a good many
nore thousands of applications a year, because the
>roblems you are facing are a concrete embodiment
if the great problem which we have before us of
narshalling the world's scientific and technical
;nowledge.

I rather feel that because of the urgency of in-
lustry and inventors and other people that are
nterested in patents, perhaps you can get at some
>f those problems a little faster than some of the
mieter, perhaps even more basic fields of science
ind technology. So, in a sense, you people are sort
)f running interference for the other people in the
ields in spite of the great achievements that have
>een made in abstracting and the analysis of docu-
nents in the other fields which are not directly re-
ated to patents.

So I am going to have a lot of fun listening to you
people and learning a lot, and it ought to be quite as
entertaining to me as it is to you. Our first paper
this morning is by the Honorable Cecil N. Jeffrey.
He is Supervising Examiner in charge of classifica-
tion and machine searching in the Australian Patent
Office. He is, incidentally, the editor of Volume I,
Number I of the Australian Patent Society Journal.
He has a graduate science degree from the Univer-
sity of Sydney. Cecil N. Jeffery will talk to us on
"Encoding and Abstracting for Universal Use."
Mr. Jeffery.

Mr. Jeffery: Thank you Mr. Moderator, ladies and
gentlemen. I am afraid that the title of my ad-
dress was chosen in rather a hurry; as a matter of
fact I am going to talk mainly about abstracting,
and a little about patent specifications and encoding
for mechanical searching.

Yesterday we discussed the subject of interna-
tional cooperation through an international classi-
fication or a concordance of classifications. I am
not sure that we reached any firm conclusion except
that it is going to be a long and difficult job to de-
velop a truly international classification or concord-
ance.

However, I think there are a few simpler steps
we could take almost at once, to minimize the
amount of repetition that occurs, particularly in
those patent offices which receive and file specifica-
tions from abroad.

Let me pose a common situation. Assume we
are in patent office B and we have just received a
large batch of specifications from patent office A.
The classification staff at B now have to incorporate
these into the search files in the right place, as
quickly as possible and with a minimum of effort.

123

Let us also assume that offices A and B do not
adopt a common classification and there is no con-
cordance between their separate domestic classifica-
tions.

Obviously someone at office B has to read the
specifications as thoroughly as possible within a lim-
ited time. Yet this has been done already, of
course, by a person at the originating office A. Our
problem is really this — to devise means whereby A
can communicate to B enough information to enable
the latter to correctly classify a specification without
having to completely read it.

I would suggest that there is a combination of
measures that could be adopted:

In the first place, an official at A has selected a
title to be printed on the first page. This title can
be very informative, e.g. "Improvements in can
openers of the triangular punch type", "Preparation
of chemical X from chemical Y." Unfortunately,
too often the title is: "A new chemical compound"
or "Improvements in electron discharge devices."
While there may be some excuse for an applicant
wanting to conceal his invention behind a vague
title while the application is pending, there seems no
justification for not adopting an informative title
when the patent is issued and public. Where, as in
my country the law prescribes that the title shall de-
scribe the invention, the law should be strictly
enforced or alternatively, the office could be empow-
ered to construct the title.

My second suggestion is that, since the classifier
or examiner at office A has been obliged to thor-
oughly analyze the specification, he is in the best
position to compose an abstract and have this
printed on the front page of the specification. Now
there are many types of abstracts and I would par-
ticularly like to draw your attention to one known
as the key word abstract. This form of abstract has
been adopted by the Standards Committee of the
American Institute of Chemical Engineers and is
described in the journal "Chemical Engineering
Progress," issues of May and June 1961, which I
have here.

In this procedure, the most important papers are
accompanied by a conventional language abstract
and also by a list of key words selected from a stand-
ard thesaurus. In addition, the key words are
often followed by letters indicating the roles of the
corresponding parts, chemicals, etc., represented by
the key word. Where a paper describes more than
one topic the key words of each topic are grouped

or "linked" together. Where a key word is specifii
it is also accompanied by the more generic tern
obtained from the thesaurus.

If such an abstract could be printed on th
specification it would enable the officer at receivinj
office B to encode the abstract into any type o
mechanized system using concept coordinatioi
with wide or narrow terms to suit the desired deptl
of analysis. Even for classification guidance, a lis
of keywords can be quite informative.

My third suggestion, which may be more diffi
cult to implement but is worth consideration, is tha
an international number should be allotted to ai
invention, in addition to the domestic serial number
This would assist in filing specifications of the sam
invention received from a number of difTeren
countries.

My final submission, for which I am indebted ft
Mr. Robert L. Birch of the Scientific Library here
is perhaps more related to the work of a librarian
but may be of use to classification officers. Th
proposal is that we should agree to have on the fron
page of the specification, standard locations for th
various particulars of the application and patenl
For example, at the top of the page would be place<
the patent number and underneath this the applica
tion number. The topmost of a list of names wouli
be that of the inventor and next one that of th
patentee. Dates could be similarly arranged ii
standard order with possibly the priority first, ap
plication next and so on. Filing and retrieval o
specifications, printed in languages unknown to th
officer, would be much simplified.

I hope that this does not sound too presumptuou
coming from the representative of a patent offic
which does not normally search through foreigi
specifications. It is more provocative than in
formative, but I hope it will help to point the wa
to fuller cooperation among our various offices.

(In preparing this talk for publication, I woul<
like to draw particular attention to the paper sub
sequently presented by Mr. M. Wolk. The pro
posal to design the drawings as an abstract is ver
pertinent and deserves close consideration. In fac
the whole question of the form and content" of th
published specification is due for reconsideration
Although the main function of the specification i
to set out a defined monopoly it is also a piece o
information and as such, has to compete for atten
tion with other forms of information. It should
like all information, be prepared in such a manne

124

that it can be entered quickly and reliably retrieved
from storage in any of the available filing systems.

C.N.J.)

Dr. Davis: Thank you very much, Mr. Jeffery.
Just to start things going, I think for titling inven-
:ions you had better get some help from headline
writers in newspapers. They have to put the story
in the headline. Is there discussion?

Mr. O'Day: I have a question on the key word
abstracting. When you have patents in ( 1 ) areas
where you have large variable structures that con-
:ain literally thousands of individual specific com-
Dounds, and, (2) patents that contain columns of
specific compounds, do you have any rules to apply
:o the key word abstracting that would allow for
;onsistency and conformity in the abstracting, or is it
strictly up to the abstractor, and do you have as
nany applications as you have people doing the
abstracting?

Wf. Jeffery: Well, sir, I am in the happy position
?f not being a chemist and therefore I cannot really
answer the question. But I would not like to go into
:oo much detail. I refer you to this journal and the
(Vork of the American Institute of Chemical Engi-
leers. I can see that there would be quite a lot of
work in some offices, but the work would be shared
and it is work that only has to be done once in one
office. I think that that is the main feature of it,
>o I do not really care how long it takes or how
detailed it might have to be.

(In response to a question by Mr. Lawlor, it was
Drought out that the thesaurus employed was the
:>ne developed at DuPont and had about 50,000 key
words. Mrs. Swanson commented that it would
oe desirable for examiners to have, on the Ameri-
:an patent, a reference to the corresponding foreign
patent number. )

Mrs. Swanson: Sir: you said that perhaps there
might not be a necessity for having many key words
in a key word abstract. Why?

Mr. Jeffery: Well, if the receiving office is an office
which only classifies, it only needs enough informa-
tion to put it in the right class. If it is an office
that encodes its specification on punched cards and
so on, of course, it needs a lot of key words. I say
it may be necessary only to have a few key words,
because I think most patent offices, for most of the
specifications they receive, would only be interested
in a very broad set of descriptors to assist in classi-
fying. The whole principle here is for the sending

office to tell the receiving office just enough infor-
mation to enable the receiving office to classify
documents.

Mr. Jancin: I am afraid that I do not fully appre-
ciate or understand the significance of the interna-
tional identity numbering system. Would you
explain that a little further please?

Mr. Jeffery: Receiving office B would receive two
specifications from different countries. It may ap-
pear that those specifications deal with the same
invention. They are from the same applicant and
they are about the same invention. But that is not
immediately apparent to receiving office B without
having to read through both and compare them.
But I have suggested, in addition to the various
application numbers, and patent numbers that ap-
pear on each specification, the use of a number
which has been allotted from a central bank of num-
bers. If that number is the same on two specifica-
tions from different countries, that means that they
describe the same invention.

Mr. Rabach: To some extent, I think what you
would like to have done is already done by the Inter-
national Classification system. We have found it
is now very easy to bring the specification to the
right man because every contracting Patent Office
is compelled to print on the specification the class
of the International Classification system, used in
multiple classification. There you have the oppor-
tunity to bring a lot of items in, just at the top of the
specification. Yesterday I pointed out, we not only
place symbols at the top of the specification, which
identify the invention, but we additionally print
symbols in lighter type which tell you of matter
which may be regarded as important, but does not
form part of the invention. And I think doing so,
we in Europe have already taken the first step in this
direction. It would be very nice if we could go
further and bring the U.S. Patent Office to the same
procedure, but I am afraid it will be a little bit diffi-
cult after the discussion we had yesterday.

Mr. Knight: We have, as most of you know, writ-
ten and published abridgements of specifications for
many years. Does Mr. Jeffery or anyone else con-
sider that it would be of any great advantage to
print tho.^.e abridgements on the opening page of the
specification, rather than as a separate publication?

Mr. Jeffery: Well, yes, Mr. Knight. The exam-
iners' files in the Australian Patent Office are pre-
pared by cutting the printed abstract out of the

125

journal and pasting it on the front of the printed
specification. It is of great use to the searchers and
I am sure it would be of great use to other offices to
have some sort of abstract or abridgement printed on
the specification.

Mr. Birch: I wonder if any thought has been given
in the Common Market countries to the publication
of a Common Patent Office Gazette — that is, a com-
bination of the various gazettes of those countries
into one? Would that be useful?

Dr. de Haan: Mr. Chairman, I might say that up
to now this question has not been raised in the Com-
mon Market as being too detailed. But I am pretty
sure if, in the future, a Common Patent will be
granted, there will be also a Common Patent
Journal.

Mr. Andrews: Mr. Jeffery has mentioned the fact
that there is a language and a key word type of ab-
stract. I would like to offer a suggestion of some-
thing that is intermediate between these two and
will not afford the criticisms from the various pat-
ent offices that a thesaurus associated with a key
word index might engender. I am referring merely
to a selected extract of terms that each patent office
considers important in its own patents. No attempt
is made to fit it into an existing thesaurus, but merely
to make a selective extract of important key words
and phrases in the patent. This would be unique
for each particular patent. It would require that
someone later organize the entire vocabulary into a
standardized one if that is necessary. But I feel
that this may overcome some of the difficulties of all
the patent office getting together and agreeing on a
thesaurus.

I might add that Mr. Rubach did propose this
key word for all the patents in a meeting we had two
years ago, and I believe that he has been very vigor-
ous in his attempts to further this action.

Dr. Davis: Thank you very much. If there is no
further discussion on this at the present time, we will
have the next paper, "Abstracting of Patents," by
the Honorable Kotaro Otani, a Chief Examiner of
the Japanese Patent Office. Mr. Otani is a me-
chanical engineer. He has been Chief Examiner
for a year, and trial examiner before that. We are
glad to have you here, sir.

Mr. Otani: Thank you, Mr. Moderator, ladies and
gentlemen. I am very happy to have an oppor-
tunity for reporting the circumstances of the exami-
nation literature in the Japanese Patent Office. As

you know, the literature arrangement is most in
portant for patent examination. And I would lili
to explain the projects that have been taken for th
past several years in the Japanese Patent Office, an
the projects that will be taken in the near futun
particularly in regard to the preparation of abstraci
of patents and other technical literature.

In Japan the number of patent applications ir
eluding utility models has been increasing year b
year, and we received over 100,000 new applicatior
last year. I just now used the term "utility moc
els". I think you are not quite so familiar with th:
term because only a few countries in the world hav
this system. But you may understand that a utilit
model is, so to speak, a small scale invention and :
different from a design.

Now, returning to the main subject, the numbe
of disposals for patents and utility models in our pal
ent office is about 100,000 per year, while we hav
over 200,000 pending applications now. Th
means that we have pending applications abou
twice the number of disposal capacity per year.

In order to dispose of such a large number c
accumulations, we are making every effort we car
The important projects which have been taken c
will be taken in the Japanese Patent Office for th
literature arrangement are as follows:

The first project is the subdivision of pater
classification. The present Table of Japanes
Patent Classification was prepared in 1948 and th
Table was revised this year. As a result, the num
ber of subdivisions of classification has been in
creased in special fields including nuclear engineer
ing, and further, the number of subdivisions in othe
fields will be increased in the future.

The second project is the preparation of ab
stracts about patents and other technical literature
In patent searching we usually use specifications c
Japanese patents including utility models an<
patents of foreign countries, technical journals o
Japan and from several foreign countries, etc.

In particular, we use as most convenient exam
ination materials, abridgements of British patents
Official gazettes of United States patents, Auszug
aus den Patentanweldungen and the like, besid
Japanese patent specifications and utility mode
specifications. Also, we have other convenien
examination materials, such as Chemical Abstracts
Science Abstracts, Nuclear Science Abstracts
Chemisches Zentralblatt and abstracts by Japanes
Information Center on Science and Technology

126

rhey are very useful to increase the efficiency of
patent examination.

Beside these abstracts, we have been making for
l few years our own abstract cards from technical
ournals of the United States and several other
:ountries in the following technical fields and use
hem for examination purposes only. The above
nentioned fields are nuclear engineering, recently
leveloped materials, polymer chemistry, petroleum
hemistry, electronics, and machine tools.

These cards are kept by the examiners in accord-
ince with the Japanese Patent Classification. The
lumber of cards already prepared is about 14,000,
nd about 6,000 cards are being prepared yearly.

These cards are serving very efficiently for patent
xamination. We are now trying to extend the
ange of abstracting to foreign patent specifications,
^nd as to these foreign patent specifications, we
/ould like to choose the specifications of the United
itates and Belgian patents next year. For this pur-
lose we have requested a budget of about 73 million
apanese yen (this corresponds to $200,000) for
naking the abstract cards for about 65,000 foreign
>atent specifications, including 50,000 United States
latents and 15,000 Belgian patents. Here we have
elected Belgian patents because in Belgium the time
rom filing to publication is very short, and there-
ore, we can know the newest techniques through
he Belgian patents. And we would like to extend
his plan to German and French patents in the
uture.

As I mentioned before, abstract cards are now
ised for examination purposes only, but we are also
danning to permit the public to use them. I think,
f they became available to the public they will
ontribute to prior searching by applicants and to
ibtain technical information.

As I mentioned, we are using many abstracts, but
imong them British abridgements are most excel-
ent and are utilized most frequently. The Official
;azettes of United States patents and Auszuge aus
len Patentanweldung disclose only a claim or
laims and the description of claims only is rather
lifficult to understand. Therefore, we consider
hat if abstracts of patents would be made and pub-
ished in the other countries besides Great Britain,
hey will be utilized efficiently in the world, for
:xamination literature as well as for technical in-
ormation. Furthermore, if even in the countries
n which English is not spoken or not used, English
ranslated abstracts could be made and published of
heir own patents, the value of abstracts would be

greatly enlarged. Therefore, I hope that the prob-
lem of patent abstracting will be considered on an
international scale. Thank you.

Dr. Davis: Thank you very much, Mr. Otani.
Discussion, please.

Mr. Purdy: Mr. Otani, you mentioned a Japanese
abstracting center, I believe. Is this a governmental
function?

Mr. Kitagawa:* No. It is a private organization.

Mr. Otani: It is between a government and private
organization.

Mr. Kitagawa: May I amplify that for you? This
is called in translation, "The Japan Information
Center for Science and Technology." It is quasi
governmental, and up until now its budgeting has
been divided between government and private in-
dustry. They have been publishing abstracts of
foreign patents as one of their functions. I believe
they also have an abstracting and translating facility
available to anybody outside of Japan.

Mr. Rotkin: To what extent does the face of the
Japanese patent bear information in other than
Japanese print? And do you contemplate any
change in that? I am thinking in line with Mr.
Jeffery's talk of the advisability in the immediate
future of having additional information on the face
of the patent to help the other patent offices in place-
ment of the particular patent in their appropriate
classification files.

Mr. Kitagawa: The first question that you ask is
what does the face of a Japanese patent contain. Is
that correct?

Mr. Rotkin: At the moment, and what is contem-
plated.

Mr. Otani: What do you mean by the face of the
patent?

Mr. Rotkin: The heading.

Mr. Otani: We have planned to express the heading
of the siibject in English, but this plan is not exactly
confined now. It is only a plan.

Mr. Kitagawa: I sort of jumped the gun on your
question. Getting back to question one. The face
is quite similar, information-wise to the U.S.
Gazette.

*(Mr. Kitagawa, from the Office of Scientific Information
Service of the National Science Foundation was kind enough
to volunteer his services to Mr. Otani, as interpreter.)

127

Mr. Rot kin: Is it entirely in Japanese print or is it
also in the Latin?

Mr. Kitagawa: Unfortunately, it is entirely in
Japanese.

Mr. Otani: We would like to consider a plan to
express it in English in the near future.

Mrs. Swanson: $200,000 was cited as the cost, but
I did not get clear what the cost was for.

Mr. Kitagawa: Rowena, this was for the purpose of
covering foreign patent literature and abstracts.

Mr. Otani: It is the cost for making abstract cards
for 50,000 United States patents and 15,000 Belgian
patents.

Dr. Davis: Is there other discussion? Otherwise, we
will get on with our program. Thank you very
much.

From the standpoint of fitting in time, we are
going to switch two papers. We are now going to
hear the first paper that is listed on your program
this afternoon and Mr. Wolk's paper will be heard
this afternoon. The Honorable Tore Oredsson,
Bureau Director of the Swedish Patent Office, is
going to talk to us on some thoughts on automatic
indexing and encoding. He is a Master of Science
in chemistry from the Technical University of Stock-
holm, chief examiner, and he has the special duty
of presenting cases in Supreme Administrative
court. Mr. Oredsson.

Mr. Oredsson: Thank you, Mr. Moderator, ladies
and gentlemen. Our main problem is how to select
out of a given collection of documents those docu-
ments that are relevant to a given topic. The ad-
vent of punched card equipment and electronic
digital computers has encouraged speculation on the
possibility of establishing an automatic data — or
reference-providing information system that would
respond to a request for data or references with an
answer providing pertinent data or references, all
this in an entirely automatic fashion and with great
speed.

The only practical solution for a reference-
providing system is to assign to each document a
clue or a set of clues, and to assign likewise to each
topic a set of topic terms in such a way that by com-
paring the set of topic terms with the set of clues, a
decision as to the relevance of the documents can
be reached. How this assignment of clues should
be done is a great problem. If manually, it will

take a long time and many people. In my opinio
we have to try seriously all means by which to obtai
these clues from the documents by some mechanic;
procedure.

Experiments with automatic indexing and encoc
ing have been made by organizations, carrying o
research in the storing and retrieval of informatioi
In an electronic computer, a comparative statist
cal analysis has been made of the frequency of wor
and concept in certain scientific and technical art
cles. The frequency established in the course of th«
analysis was either the absolute frequency of a wor
in a given document or the ratio of the relative fr<
quency of the word in the document to its relatrv
frequency in the language in general. The signif
cance of high frequency words and concepts for th
identification and encoding of such articles has aftei
wards been estimated.

The methods used and the results obtaine
thereby have been criticized. Among others, Pre
fessor Yehoshua Bar-Hillel, Jerusalem, put forwar
criticism in a very interesting and constructive pape
called "Some Theoretical Aspects of the Mechanize
tion of Literature Searching." I am going to rel
to a great extent on his paper. Criticism is desii
able and necessary since it gives incentive for in
provement of the methods used. However, yo
may think that to some extent the criticism has gon
too far. It has been shown that there are deficier
cies in the new methods, and these have been foun
so serious that the methods have been considere-
useless. However, one seems to have forgotten tha
the present methods for information retrievel ar
also very defective.

Although a large amount of work is being laii
down on conventional classification systems, sue!
as the American and International Classification
system, these systems are still imperfect. Th
classification of new search material — for instance
new patents — according to a conventional classifi
cation system may be very difficult, and this is espe
cially true if the system is very subdivided. Differ
ent results may be obtained depending on th
person who is classifying. Furthermore, it is some
times difficult to know which class, sub-class, etc,
has to be searched. The most important one ma
be overlooked. Finally, the examiners may hav
very different points of view as to the relevance c
a certain document. A document, which examine
A considers to be a reference, may be completel
ignored by examiner B.

128

Let us, therefore, set our aim on a practical target,
ind be satisfied with a system that just works more
atisfactorily than the existing ones on the average
or a given group of users — namely, the examiners.

In my opinion, there is no reason a priori to dis-
[ualify a search material or procedure constructed
>y automatic indexing and encoding. On the con-
rary, one should find out which are the problems
ti automatic indexing and subsequent searching by
neans of stored indexes obtained through auto-
matic indexing. Then one should analyze and
valuate those problems as thoroughly as possible.
Vhat is ultimately decisive is how the reference list
ius obtained will appear as compared with those
btained by competent humans with regard to
uality and cost.

I would like to emphasize at this point that I
ave had no possibility of making any practical ex-
eriments. All I have done is to read certain
terature in this field. Against that background I
'ould like to try to give you some of the problems,
o start with, I will disregard the difficulties due
d the fact that different languages are used,
'urthermore, I assume that the machine can be
lpplied at no or very little extra cost with a
lachine searchable version of the original text.
^he sequence in which I mention the problems is
erfectly arbitrary and of no importance.

1. The first point would be the problems of
^nonyms, comprising also all the phrases of the
ame meaning as the individual words. As ex-
mples of synonyms, we could take the concepts
distillation", "rectification", "evaporation" — and
s an example of a synonym phrase, that "a liquor
i distilled or vaporized." This problem is well
nown; to cope with it, it has been suggested that
synonym dictionary be used — for instance, Roget's
'hesaurus. In a subcommittee for documentation
ppointed by the German Association for Docu-
mentation so-called concept-lists are being estab-
shed in different technical fields, if I am rightly
lformed. Also, in the Australian Patent Office
mere have been developed "descriptors" of the same
ind. And I believe that, for instance, ASTIA
as something of the same sort. Combining the
laterial thus already established with the material
btained by the frequency analysis which I men-
;oned in the beginning, a valuable material with
diich to start the creation of a thesaurus of the kind
have in mind may be at hand. Such a dictionary
rould be of help in the reformation of the original

topic when a literature search based on the original
formulation did not yield satisfactory results. I
shall be coming back to this presently.

It might also be of use for the authors of new
articles and patent specifications. Perhaps it
would contribute to the forming of an interna-
tionally standardized technical and scientific lan-
guage, which, in its turn, would simplify the coding.

The dictionary should be arranged alphabetically
by term, as well as numerically by code. All the
synonyms should be collected under one code num-
ber in the numerical list, but each one should be
found in its alphabetical place in the alphabetical
list.

I have been asking myself, if it might perhaps be
of advantage if the terms could be grouped to-
gether — for example, under the following headings :
Terms pertaining to construction, to function, to
use, to manufacture and fabrication, and to tech-
nical characteristics. Which group to use will de-
pend on the technical field in question, and the
manner in which, according to experience, the
examiner will proceed in searching that field.
Such grouping should, to my way of seeing it, facili-
tate the subsequent information retrieval. Of
course, there are many points of doubt about a
grouping of this kind. I just mention it to see if
perhaps we might learn something by discussing it.

2. The second point would be the problem of
homonyms — that is, when one and the same word
has several different meanings. Take, for instance,
the word spring, which may be the metal spiral
supporting the chair you are sitting in or the season
of cherry blossoms or even a fountain, and a good
many other things as well. Certain languages have
very many homonyms, and this makes for difficulty,
but the difficulties should not be exaggerated; within
limited technical fields, you will get homonyms only
to a limited extent.

3. The third point would be the problem of nega-
tives and of low frequency concepts of importance
for the identification of documents. Even if you do
not take any special measures to master this prob-
lem, the quality of results you would get in retriev-
ing information would, in practice, not be appreci-
ably reduced thereby.

4. The fourth point would be the problem of
different forms of one and the same concept — for
instance, difference, different, differently, differenti-
ate, differential. Insofar as they are synonyms, the
problem is already mentioned in the foregoing. And

129

to the extent that the meaning is dependent on the
context, the problem belongs rather to the point
which I am just about to talk of — that's the fifth.

5. Thus, the fifth point would be the problem of
dependence of words or concepts on the foregoing
and subsequent concepts — that is, the context. "A
nuclear reactor", thus, is something altogether
different from the words "a reactor nucleus" at least
in the Swedish language. "The cooling of water"
is not the same as "cooling with water", etc. This
is a problem of syntax, analogous with the problem
you would encounter in machine translation. When
responding to a certain inquiry, all indexing systems
would, for this reason, produce a number of refer-
ences of no relevance to the inquiry, so-called false
drops. However, this fact is not a cause for any
special worry, if the number of false drops does not
become too large. One solution might be to con-
sider word pairs or, in some cases, phrases. Methods
for encoding relationships in context by expanding
the codes of these index terms to contain syntactic
"role indicators", "interfixes", etc., have been pro-
posed by, among others, the U.S. Patent Office
group. These methods make the coding, the re-
trieval operations and the rephrasing of the search
question in preparation for the retrieval, more com-
plex. The only way to prove, if it is necessary to
encode relationships, is to make comparative tests
without as well as with such interfixings.

6. The sixth point would be the problem of supe-
rior and subordinate concepts, and of concepts or
combinations of concepts that partly cover one an-
other or overlap. If you want to know, for instance,
what is known about methods for exterminating
insects with chlorinated hydrocarbons, a paper on
extermination of flies with DDT would be of inter-
est. This problem was stated in the introduction
to the papers published in the area 6 of the Inter-
national Conference on Scientific Information,
1958, here in Washington, in the following way:
"Concept of relatedness or connectivity of docu-
ments in a collection, along with a suitable metric
to define the degree of relatedness." Up to now,
nobody seems to have solved this problem satisfac-
torily, even though certain theories have been
brought forward.

7. From this I wish to run over to another diffi-
cult problem, the problem of references that are
relevant even if they are not deadly citations; for a
document is not necessarily either relevant to a given
research topic or irrelevant to it — it is rather the
degree of the document's relevance which is im-

portant. This last problem is nothing special fc
mechanized searching. As I mentioned earlier, th|
same problem exists in manual searching. It ii,
however, a real problem with which we have t
cope.

In summary, we can notice several problems i
mechanized indexing and encoding and in mecha
nized searching of files prepared in this way. Tr
most difficult of these are, in my opinion, first t\
problem with generic and subgeneric concepts an*
concepts or combination of concepts which parti
cover one another, and secondly, the problem 1
relevance. I would like to find out if the solutio:
to these problems might be not to use the compute
in the way it is now generally used. When a com!
puter is now used for computing, it is best for eco
nomical reasons to program the solution of a prob
lem and to have the computer work on it withou
any human intervention until the final solution i
pointed out. However, the best way to make us
of an electronic computer for information retrieve
might be, as Professor Bar-Hillel suggests, some kirn
of intermittent employment. The decision as to th
changes to be made in the original set of topi
terms in order to get additional references, shouL
not be left to the computer. Whether a new searc
should be conducted on an index obtained from th
original one by omitting one of the topic terms b
substituting for one of the topic terms some othe
term, or whether the request is to be rephrased altc
gether, all that is probably best decided upon by tl
examiner.

The statistical analysis of the word and concep
frequency, which I have already mentioned, should
as I see it, be riveted on the specification as a whol
and not merely on the claims. By doing so, on
would find out if an indexing of the claims on|
would, in certain cases, be sufficient. This wouL
be the ca-~e particularly with regard to patents fron
countries that require a certain conformity betweer
description and claims.

The analysis should be directed to fields wher|
similar work has been or is being undertaken in dii
ferent countries. As an example of a field, whic!
is thus the subject of an intensive study in man
places, I might mention the technology of atomi]
energy. In the German Committee for Documenj
tation, there is at present being put together
concept list relating to atomic reactor technique. L
CETIS, an organization within the EURATOM
work is right now being conducted with a view to th
setting up of a list of "descriptors" in this fielc

130

'erhaps lastly, the Australian and British Patent
)ffices may have access to lists of "descriptors" and
iractical experience as to Permutation indexing in
be field of atomic energy.

)r. Davis: Thank you very much Mr. Oredsson.
s there any Discussion?

4r. Fairthorne: So many very important questions
lave been raised here, in fact all of the important
uestions I think people have been trying to solve for
ome fifteen years, that one deserves comment on all
f it. We have had a very good summary, from
/hat I have heard. One thing I want, however, to
uggest to anybody who is dealing with some of these
nethods, roughly speaking the first time, is not to
onsider whether they work. As a matter of fact,
hey do, mainly because group-fact language does
/ork, which is a fact you might overlook if you have
ead Dr. Bar-HillePs paper. You are surprised if
ou understand any words at all by the time you
lave got through. But my opinion about Dr. Bar-
lillel's paper has been expressed at length over
here.

You see, if you take the question of statistical
nalysis of the text, which, by the way, runs right
>ack, and I would like to give the name of the per-
on who really did the first work on it in modern
imes, which was Dr. Richens, of Cambridge, Eng-
and, I suppose I shall have to say, right back at the
:nd of the war. I believe it is hampered solely, of
purse, as most people have a lack of equipment for
(his very, very difficult task. You must use the
:ntire text. There is no doubt about it. Other-
vise you get in the mess of taking an abstract which
is what somebody else thought about the text, as
)eing the equivalent to the text. The only reason
he people do it is because it is very difficult to han-
lle complete texts. It is extremely expensive to
ranscribe it and examine it and then do analyses.
Generally speaking, the only things that are done
ire Shakespeare, the Bible, and Sir Thomas
\quinas. But you cannot do that sort of thing to
iverybody.

Now, very exaggerated claims have been made for
tatistical analysis and other things. And it is just
as well to think, not whether these things will work,
j)ut how far they can work, and whether it is worth-
while pushing them any further. Now it is quite
:lear that statistical analysis, as it stands, will only
10 up as far as an anagram of what you are dealing
with. All anagrams are the same. Now it is ter-
"ibly obvious if you are given a lot of words and the

number of times you are allowed to use them, you
can write a great many articles doing just that. And
they are not all the same. Nor would they be
indexed all the same, nor would they always be the
same claim. This obvious fact seems to have been
overlooked.

Nevertheless if you can take the anagram of a
paper, going up to a certain degree, that is a very
good way of doing it. But do not push it too far
because it becomes ridiculous. And this goes for
every other method including those that I use myself.

So I would just conclude, without going into
details, do try the thing. But there are obvious
limits beyond which you should not push precision
on these things, when it ceases to be without mean-
ing, in other words, as clearly an anagram will.
And also remember, that if you are looking for any-
thing that has large public agreement as to what
the words or phrases mean, it cannot be very spe-
cific. And if it has to be very specific, it will not
be either very widespread nor will it be very stable
in time, because where people are most specific,
they are changing their interests most often. And
this is a great trouble, and the way one hopes to deal
with it, is by having thesauruses. But such matters
in concordances would have to be revised almost
daily. For what we are dealing with is how people
use words. We are not dealing immediately with
why they do it. So machines at least can keep our
things up to date day by day. But you cannot have
both stability of your classifications and specificity.
That is that. I do not see any way out of it. It is
exactly the same thing Compton was telling you.
You can have one or you can have the other, but you
can not have both.

Mrs. Swanson: There is a collection of about ten
years of patents in the chemistry field which has been
indexed very roughly by a commercial organization.
They have prepared a chemical patent index which
I am sure many people are familiar with. Since I
had difficulty in examining chemical patents at one
time, I had often wondered what the utility of this
product would be in at least getting a broad cut
to the arts that might be significant in handling an
application. And since I have kept this in me for
a long time, and this seems to be a forum in which
it can be mentioned, I wish to suggest that this prod-
uct be given a test to see whether it could actually
be useful to the patent examiner in the field which
is covered. It ties in a bit, I think, with the
speaker's comments on concept. The indexing is

131

not standardized. It is a very rough job, but it may
very well be useful, and anything which would cut
down the manual operations of looking through
piles and piles of patents would appear to be a bene-
ficial thing.

Mr. Van Waasbergen: I would like to reply to what
the lady said. The people of this company ap-
proached us and we looked into it and we came to
the conclusion that for our work it was a little bit
too rough. It could be used as an introduction to
the study of the patent application, but for that in-
troduction we found it too expensive.

Mr. Kalikow: I would just like to congratulate Mr.
Oredsson on his talk. I though it was very provoca-
tive. It is completely consistent with the exper-
ience that we have had in the General Electric
Company in patent searching by mechanical means.
We need to watch the crossover point between man
and machine, and be able to use the machine to the
extent that it will give us advantage and be able to
have pertinency factors built into it. Then when
you get your retrieval you can take a predetermined
amount of overpull whereby the human mind can
then make a further decision as to how much fur-
ther it wishes to go either in reliance on the machine
or going back to other methods of searching. Using
this machine as a practical tool, and these systems
and the programs that he has mentioned as prac-
tical tools, actually works very well.

Mr. Pfeffer: I have one brief comment for Mr.
Oredsson with respect to his cavalier dismissal of
negative limitations. At least it has been our ex-
perience, in our practice in this country, that nega-
tive limitations are nothing to be ignored, especially
with respect to the art dealing with compositions of
matter. Now I understand that this is not a subject
for patentability in many patent offices, but it is in
this country. Quite frequently we find that the
best way of expressing the invention is by the use of
some sort of negative limitation — for example, a
non-aqueous composition or a non-chlorinated azo
dye, where the absence of a particular thing is the
gist of the invention for the specific result desired.
For this purpose the search is a very difficult one
for the examiner, because he must find a statement
somewhere in a specification which teaches this
negative implication. He cannot merely rely upon
the fact that some combination of elements has
neglected to mention this fact. So this for us be-
comes quite an important concept, and we cannot
ignore it.

132

Mr. Purdy: Would you mind sir repeating brief!
what are your conclusions regarding the differer
forms of the same concept, such as difference, diffei
entiate, differential, and so forth.

Mr. Oredsson: Well I went over that subject rathe
quickly. I am afraid I said that to some exter
they could be considered as synonyms, and in the
way you take care of it in the thesaurus to whic
Mr. Fairthorne was referring. Otherwise, th
meaning is dependent upon the context. I do nc
know what the people say who have experience i
this field.

Mr. Kalikow: I can give you some experience in th
field. We use what we call a number of differenc
thesauruses. One is the root word thesaurus, wher
all of the words having different endings of the sam
kind of word, the same kind of meaning, lik
saturate, saturating, saturated, saturates, and s
forth, would all come under one word, which woul
be the root word thesaurus. Then we move bac
from that into the representative word thesauruf
which is about equivalent to your Roget's The
saurus. Then we have that divided into a comple
word thesaurus and a basic word thesaurus. Yo
have the complex word thesaurus which is one tha
has all of the words that are factorable into thei
component meanings, and then you take the word
in that thesaurus and you break them down in term
of the basic word thesaurus that contains words tha
are very difficult to factor any further. With thi
kind of a hierarchy of thesaura, you are able to mak
these equivalences.

Mr. Rotkin: The speaker's comments about th
analysis of language sounded familiar but it did no
quite fit in until Mr. Kalikow spoke. I am wonder
ing if possibly we have missed a bet by not turnin
to some of the work that has been done recently h
translating machines. What you said sounde<
exactly the way the recent papers in the translatin
area have approached the problem of getting wha
they call information from one language int
another.

Mr. Kalikow: I think one of the most encouragin
things in this whole field is the way people fror.
many, many different viewpoints are coming to th
same conclusion.

Mr. Reznek: I am wondering in this discussion c
thesauruses whether we are not concerned with ere
ating sort of a general thesaurus which would b
gigantic. Actually what is needed is a thesauri;

hat relates to different arts. Words have different
neanings within different frames of reference. We
nentioned spring. Spring in one is a mechanical
concept, another a hydraulic concept, and the third
i biological concept. I think we are too much con-
erned with the meaning of a word which not only
las a different frame of reference in an art, but
'aries from person to person. And you must keep
hat concept in mind. I want to give an example.

was classifying static structures and I used the word
haft for the generic concept of columns, beams,
tuds, and so forth. There was not any satisfactory
vord. When one of my colleagues began to work
m the class he said "What is this shaft business?"
ind I explained it. He said "Well, to me a shaft
s a part of a machine, as a rotating shaft." The
)icture presented to him was entirely different than
hat to me. That is something that to me is insolu-
)le. We can not have an individual thesaurus for
:very person. But I think we can have a thesaurus
or different arts, and of course we have to deter-
nine the extent of the art.

A recess for lunch was called by the Moderator
ind the meeting reconvened at 2:00 P.M.

\fternoon

Chairman Fay: There has been some indication on
he part of some of you that you would like to see
he Patent Office. I know some of the foreign
'isitors have been looking at our operations and
:omparing notes. One of the men from Canada
vas there today, looking at certain phases of our
vork and comparing against problems that he has.
some of you from industry or from universities or
esearch centers may never have seen the Patent
Dffice. If it is your desire we would be very pleased
o arrange to have you see any parts of the office that
'ou would like. The visitors from abroad have
seen invited to a meeting at the Commissioner's
sffice Thursday at 9:00. Now, if anybody else
vould like to come down either Thursday or Friday
ind have a tour of the office or look at specific parts
)f it or see how we are organized, or look at the ex-
imining function or our mechanical search division,
f you will indicate your desires on your pads, Mr.
jorecki will collect them and we will attempt to
irrange such things as we are able to do. As far as
know, all of Thursday and Friday, we can arrange
or tours or discussions with our people, and if there
s anything else that we can do for you while you are
lere in Washington, please ask us and we will do
)ur best to comply. Are there any questions? I

also want to say that the Commissioner hopes that
the Workshop is going well, and that you are enjoy-
ing it. If we can do anything more on this let us
know. I appreciate Dr. Davis' helping this morn-
ing. Thank you.

Dr. Davis: Thank you, Commissioner Fay. We will
now have the paper by Morris O. Wolk, Assistant
Chief of Division 43, the United States Patent
Office, which includes bleaching, dyeing, preserv-
ing and medicine. I hope the medicine and "dye-
ing" do not have anything to do with each other.
He is a chemical engineer and has a law degree as
well. The title is "Information Display for Encod-
ing and Post-Retrieval Benefits." I hope somebody
tells me about what this encoding means some day.
Maybe he will. Mr. Wolk.

Mr. Wolk: This meeting has been labeled a "work-
shop" on information retrieval. This talk is
planned, in part, as a workshop demonstration on
information display.

In the folder which each of you has received
there are numerous drawings which we will be re-
ferring to. This is obviously necessary since our
subject is information display or presentation (the
latter term being now popular and equally appli-
cable). So I will refer to an item in your packet
and expect that you will have it before you for in-
spection as it is being discussed.

Some of the material may be regarded now as
being well proved. I refer, for example, to your
xerographed copies of the drawings of the four
patents as issued to Webber, Silver, Rogers and
Douglas. The well-proved part is the use of a
single standard pattern of drawing in chemical
process cases, which we will hereinafter draw atten-
tion to as designatable as being of modular design.
There will be subsequent comment regarding these.

Other parts of the materials supplied are new, as
far as is known; involve extension to mechanical
patents of ideas on information display previously
urged for chemical patents, and may generate ac-
tive discussion from our friends primarily interested
in mechanical patents. For example, your packet
material on the drawings, or modified drawings, on
the Veit and the Scott patents, are respectively, on
a method of forming riveted joints and on a doctor-
blade.

A suggestion will be made that Chemical Ab-
stracts or British Chemical Abstracts or any other
such abstract journal could well use the drawings
of patents such as those of Webber, Silver, Rogers

133

Feb. 28, 1956 c. s. webber 2,736,718

PHENOLIC RESIN PRODUCTS AND THEIR PRODUCTION
Filed Aug. 2, 1952

C 6 H.

Agitate

Constantly,
Heot to

70-100° C
in I- 2 hours
and hold
for 2-6 hours

I mol

or other monohydric phenol, e.g., cresols,
dimethyl phenols, nitrophenols, chlorophenols,
and mixtures thereof, etc.
os C. P. solid or melt or commercial product.

HCHO

or other oldehyde, e.g. .formaldehyde
37%oqueous polymers, acetaldehyde, furfural,
mixtures thereof, etc.

Ungelled Phenolic
Resin Syrup

Water ond
Other Volatiles

m disti

off

l.4to3mols or os gaseous HCHO or in

alcohol solution.

or hydroxide of
K,Li,Ca,Ba,Sr,Mg,
Zn.AI, NH 5 ;
Carbonate of K.Li.No, Sr
amines, quarternary
ammonium bases etc.

or as solid

Vacuum

"A"or"B"Stage Phenolic
Resin Solution having
Solids Content of 60 to 90%

Mix to

Homogeneous

Solution

aqueous solution
of 60 to 90 % ■""
solids content

• films coated from the solution
require at leost 3 hours
to cure ot 190° F

C 6 H 5 ONa

ltol5%,ond best 2-8%
based on weight of the
resin In the solution

RAPID CURING
PHENOLIC R ESIN

Milms

os melt in water
of crystallization
or as aqueous or
alcohol solution

coated from
the solution may be
cured ot I90» F
in I to 2 hours

■ or the Na,K,Li,Ca,Ba,
or Sr, phenolate of
phenol
resorcinol
hydroquinone
chlorophenols
olkylphenols
nitrophenols
carboxyphenols
orylphenols
aryloxyphenols
ary lalkylphenols,
mixtures, etc.

INVENTOR

Charles S. Webber

BY CUt^ ^ &/~M

Figure 1 ATTORNEY

134

Nov. 18, 1958

I_ H. SILVER

ANTIBIOTIC DUSTS
Filed Nov. 12, 1957

2,861,024

DIHYDROSTREPTOMYCIN
SULFATE (DHS-S)

Dust -like
material

agitate
in mixing
vessel

fine
beodlets

milled < 50 **
90 %< 10 ^

or other antibiotic of fine part size

Streptomycin ■»■ its salt*

Dihydrostreptomycin * its salts

Neomycin

Aureomycin

Erythromycin

Terramyein

Tetracycline

4 - amino - 3- Isoxazolldone

or combination* of the**

COATING ANTIBIOTIC with

SiO,

< 15^

< 2^,pref.

05 to 20% by weight

of antibiotic
I to 5% preferably

.Si0 2 or other anti- caking agent
such as water insoluble soap,
vegetable flour, water insoluble
inorganic magnesium comp.

DHS-S

with

COATED

SiO t

/may
K omit

CaCOj

add

Co CO,

DHS-S COATED

with Si0 2

plus Ca C0 3

Dusting
Apparatus

"G>

Non- clogging and non-adhering
coated particles of antibiotic*

or other inert anti- caking agent, as
Storch

Ca phosphate
Ca chloride

OPTIONAL , /'/ comp maybe
exposed to 'excessive moisture

These SiOj coated particles
are easily broken down to
realize the fine particles of
antibiotic* In their original state.

05 to 2%

-Hand or mechanical dust gun
for dust application, as to
plants, animal*, etc.

DUST OF DHS-S

fine part iclt i
in o r i q i r> a I State

Nozzle of dusting apparatus

has a restriction —

screen or orifice

opening 0.125" to 0.15" dlam

pref. % 4 ' dlam.

INVENTOR.

Leonard H. Silver

ATTORNEY

Figure 2

645803 0-63-5

135

Sept. 8, 1959

S. M. ROGERS

Na.0CL-H 2 G> COTTON CLOTH BLEACHING
Filed July 3. 1957

2,903,327

COTTON CLOTH

Caustic Soda

^ r immerse in bath
10 sec. to 3min.

Aqueous Bath
5-45°C.
pH 9.0-12

Cl 2

odd, to make up

ftath 05-3 grams
NoOCI per liter

NaOCI
(Aq.)

free of Carbonate ions

COTTON CLOTH
NaOCI Impregnated

Squeeze to 70-150% saturation

Stand 5° to 45°C. 5 to 45 minutes

COTTON CLOTH
NaOCI

Aqueous Bath

0-80°C
pH 10.0-120

partially bleached
incompletely exhausted

immerse in bath

2.0 to 12.0 grams
per liter

Squeeze

to 70-150%
saturation

Stand

Wash

75° to!50°C ^ »o2 hours

water (distilled water may be employed, but
it not essential)

BLEACHED
COTTON CLOTH

INVENTOR
STUART M.ROGERS
BY

ATTORNEY

Figure 3

136

Aug. 8, 1961 J. F. DOUGLAS ETAL 2,995,493

FICUS ENZYME AND HYDROPHILIC CARRIER COMPOSITION
Filed April 17, 1957

Raw Fig Sap

Mix

Filter
Cake

discard

.from a ficus latex,
of the group
laurifolia
glabrata, and
carica

Filter Aid

Filter

Filtrate

Liquid

d iscard

-or sodium chloride, magnesium sulfate,
sodium sulfate, potassium sulfate,
sodium acetate , mono-potassium di-hydrogen
phosphate, or mixtures thereof, etc.

(NH 4 ) 2 S0 4
Concentration
30 to40% of saturation

Hold

for 24 hours

Centrifuge

Solid Material

i — or citric acid with sodium citrate,
acetic acid with sodium acetate,
or the equivalent systems with other
alkali or the like cations, etc.

Form
Solution

may be

packaged and

sterili zed

Di-sodium phosphate

and citric acid buffer
(aq)

Dialyze

Filter

Lyophilize

Dry Powder Enzyme

shortly before use

Mix

Carrier

(sterile)

such as

water

Enzyme & Carrier
Composition
containing
I to 10% of Enzyme

PH
4 to 7

or a mixture of
2.0 gms. algin

4.0 gms. carboxymethylcellulose
7. 5 gms propylene glycol
22.5 gms. glycerine
I 4 gms distilled water plus
di-sodium phosphate and citric acid buffer (aq).
Urea may be included; or a mixture of

1.72 gms. polyvinylpyrrolidone ( M.W. about 40,000)

2.58 gms. carboxymethylcellulose

0.30 gms. calcium chloride

1.60 gms. solid non-ionic water soluble

condensate of ethylene oxide and propylene
oxide of a molecular weight above about
7500 and sterilized by moist heat 110
parts of water; or a mixture of

6.00 gms. glycerine
2.00gms. propylene glycol
1.13 gms. benzyl alcohol
90.87 gms 0.1 m. citrate buffer (pH 5.1 ) ■,
or other hydrophilic carrier base of
10,000 to 50,000 cps. viscosity, etc.

Figure 4

137

138

JAN. II, 1955

L. H. CONOVER

Tetracycline

Filed Oc<. 9, 1953

EX.1 and General Disclosure

CHLORTETRACYCLINE

The Starting Materia I (React ant)
Preferred is chlortetracycline itself,
i.e.,the amphoteric compound free
from metallic ion or acid

The solvent must be relatively
inert to the reactant and to the
product formed.

In general, the solvent or mix-
ture of solvents should be rel-
atively dry. A small amount
of moisture does not interfere,
seriously with the ne";
action.

Aureomycin; or salts of same

pure, 4-8 grams
added if start-

2,699,054

2 sheets, sheet I

,or other slightly soluble, mild base
such as

1BARIUM CARBONATE

ing material is _

an acid salt, such as chlortetracycline hydrochloride,
to neutralize the HCI formed during the reaction
,to suspend the

calcium carbonate,
strontium carbonate

chlortetracycline

METHANOL

. to completely

ISOLUTION

dissolve prod-
uct

DIOXANE

anhydrous

Starting material conc'n
1% to about 10%
(on weight of sol'n ) |
satisfactory for the reaction

Higher or lower conc'ns useful.

Conventional

Hydrogenation

Apparatus

Room Temp, or

up to 50° C
Usually I to 12
hours, for
Reaction ternp.,time, [appreciable yield
press., conc'n of antibiotic, type of
catalyst, and like operating cond-
itions are interdependent.

off

IColalystfr -

boiling

- (DToxonel
boiling

wash 1

or mixture of methanol and dioxane,
100 ml or other organic solvent, such as

.RTETRACYC. ^""^ °P°lar solvents, as
ethanoL isopropanol, etc.

certain ethers are useful, e.g.
cellosolve, tetrahydrofurane,
dimethoxyethane
or other form of palladium catalyst, such as
5% finely divided palladium itself

other catalysts may be used, such as
platinum, or
Raney nickel
0-5 grams Necessary conc'n. of catalyst is at least

5% of weight of antibiotic used, and
usually not greater than 100%.

CATALYTIC HYDROGENATION TO DECHLORINATE

as hydrogen-

ation catalyst
MIXTUREI

PALLADIUM

ON
CHARCOAL

add, while

agitating
mixture

HYDROGEN

50 p. s.i
In general, after approx. one molar
quantity of hydrogen absorbed by
reaction mixture, rate of absorp-
tion tends to diminish or cease.
Product may then be recovered.

FTTfeTl

Atmospheric pressure of hydrogen usable.
Slightly elevated pressure speeds reaction
Pressure up to 200 p. s.i. suitable
5 to 50 p.s.i. generally suitable

SOLUTION
wttn crude
ANTIBIOTIC

|Catalyst|

SOLUTION
with crude
ANTIBIOTIC

Advisable to
avoid elevated
temperatures
during recovery
of the product;
slightly higher
thon room tem|
not deleterious

|Liquor| <

alternative

Concentrate!

» IDIETHYLETHERl

or other non-sot vent

for the product

preferably an organic

non-polar solvent, as

pentane benzene

toluene chloroform

diethylether

THE HYDROGENATION reduces the chlortetracycline
to a compound lacking in non-ionic (i.e., carbon-
attached) chlorine.

During the reduction a molecule of HCI is formed
This combines with the basic function of the mol-
ecule to form the hydrochloride of the new anti-
biotic, tetracycline (unless some more basic
material is present to react with the HCI ).

When treated with silver
nitrate solution, gave a
positive test for chloride ion.
Possessed a strongly acidic reaction, demonstrating release of the non -ionic
Potency (by bioassay) chlorine in the form of hydrogen chloride

580 mcg./mg. (approximately)

using oxytetra eye line as standard at
potency of 1,000 mcg./mg.

Concentratel under vacuum, at room temperature

Freeze, Evocuotel i.e., freeze- dry

TETRACYCLINE
HYDROCHLORIDE

TETRACYCLINE
HYDROCHLORIDE! 31 grams

Amorphous, bright yellow, solid product, soluble in water
The hydrochloride is formed if no basic material is included
in the reaction mixture, where pure chlortetracycline
(i.e., not its hydrochloride or other salt) is the starting material.

| Water I to rediss olve, but in oq
, to corefully

"neutralize to pH
4-5

TETRACYCLINE
(aq.)

TrTDAwri iKic (Tetracycline base(free base)
I 1 1 KAL-TL-LINt Appreciable solubility in »

or other alkali metal hydroxide, as most conven
JT7 — _ u — <T3ilute ienl for adjusting the pH of the aq. solution
IMOUH (oq-)| sol'n. until sol'n is slightly acidic (pH of approx.
4-0- 50 preferred)-

This converts the tetracycline hydrochloride
water to tetracycline ("tetracycline perse")

as by evapor
Dry | ation, or other
wise careful- —
ly remov-
ing the
water

■ to extract the
* amphoteric
tetracycline

BUTANOL

or other solvent, such as

ether
or like water-immiscible organic solvent

AQUEOUS
EXTRACT

BUTANOL
EXTRACT

SOLVENT EXTRACTION FROM AQUEOUS SOLUTION

BUTANOL
EXTRACT

COLLECTED

BUTANOL

EXTRACTS

I BUTANOL) ■tt 8f «ES£«'y

INVENTOR.

TETRACYCLINE

Base, as solid
material

The
Base

I Concentratel to small volume

I TETRACYCLINE] Cryst alline base

L3 to re-

J/^rystallize
v ,H

TOLUENE

ighly purified, crystalline, amphoteric product. Anhydrous.

TFTRAPYPI IMF Esp. easy to obtain when starting material is pure chlortetracycline.
JJl_l_!_. v - ,L -"\£JThis is the tetracycline of J.A.C.S.Vol. 74. pp 4976,-7, Oct. 5, 1952

Figure 5

JAN. 1 1, 1955

Filed Oct. 9, 1953

L. H. CONOVER

TETRACYCLINE

2,699,054

2 sheets, sheet 2

TETRACYCLINE

Advontages as an antimicrobial
agent when compared to the par-
ent chlortetracycline compound,
e.g., more stable in aqueous
preparations particularly
at sligrnly bas ic pH's

The highly puritied, recrystallized, anhydrous product (see bottom of
m. 170- I75°C, with decomposition sheet I)

optical rotation [p]p <-5 = — 239° when dissolved in methanol at a
on titration in 50% aq. dimethylformamide (Concentration of 1%

two acid constants observed, pKp 8-3, 10-2
Analysis: C, 59 35%; H, 541%; N, 615%
Calculated: C, 59-45%; H, 5 44%; N, 6 -31%, for C22H24N2O3
Equivalent wt. by titration - 22 7
Calculated molecular wt. = 444
Me thanolic solution 00 1 N in Ultraviolet absorption maxima

Hydrogen chloride

S odium tiydr oxide

about 270 and 362 mu

at 245, 265 and 362 m^j

I TETRACYCLINE! |TETRACYCLINE| S r e ,e,racydine is an amph0,eric

The acid or base

to dis-

solve

jWATER|

Since the basic

particularly strong,

a strong

the form

Salts of the

chlortetracycline one

tetracycline and by

to form

acid addh
salt

app. I equiK

[HCil

should be chosen
Thus, if the salt is

an acid
If the antibiotic is

the toxic

compound, just as are chlortetra
cychne and oxytetracycline,

a variety of salts may be pre-
pared from the antibiotic with
various acids or bases

. to dissolve

-do.

METHANOL

or other suitable solvent
and acidic groups of the antibiotic are not

acid or a strong base is preferably used in
ation of stable salts of the antibiotic,
same type asthose previously prepared from
oxytetracycline may also be prepared from
the same processes.

approximately one equivalent

1 to form acid

addition salt

"acid salt"

of tetracycline

HYDROCHLORIC
ACID

or other preferred
mineral acid whi ch is
strong enough to form stable
tetracycline salts, such as
Sulfuric acid
Nitric acid
Phosphoric acid

keeping in mind the use to which the salt is
to be administered to humans or animals int
or base is chosen which is relatively non-
to be used for treatment of, for instance,
ity of the acid or base used for forming the

--da

-do.

a pp. I eg.

NoOHl <9

1 m

to be put.

ernally,

toxic.

plant dis |eases,

salt is not

do (ditto) refers to
corresponding entry
horizontally to left

approx. one equivalent

.to form

metal salt
of tetracy-
cline

SODIUM
HYDROXIDE

or other
strong base such as
an other alkali metal base
Potassium OH
Lithium OH
etc., or
alkaline earth base
Calcium OH
Barium OH
etc.

nearly as important.

TETRACYCLINE

HYDROCHLORIDE

(aq.)

solution

LS^rafe

carefully

TETRACYCLINE

HYDROCHLORIDE

Solid salt

TETRACYCLINE

HYDROCHLORIDE

(in methanol)

Meth
anol

off

Evap-
orate

carefully

TETRACYCLINE
HYDROCHLORIDE

Solid salt

while in

frozen state

[Liquor

off

solution

NaSALT OF
TETRACYCLINE]
(aq.)

solution

--do.

Na SALT OF
TETRACYCLINE

.to precipitate

the salt

ETHER

Solid salt

or other non-solvent
for the salt

TETRACYCLINE
HYDROCHLORIDE

Na SALT OF

TETRACYCLINE

(in methanol)

--do

SALT

TETRAC

Solid salt
-do.
-do

do-

solution

—do.

NaSALT OF
TETRACYCLINE

Na SALT OF
TETRACYCLINE

Solid salt

or other metal
salt, correspond
ing to base
used

INVENTOR.

or other acid
addition salt
correspond-
Solid salt Solid salt ing to acid used

Tetracycline and tetracycline salts possess a high order of activity against a variety of micro-
organisms including both gram - pos itive a na gram- negative species.

A serial dilution procedure gives values of this act iv ity in vitro as foil ows:
Species & Minimum Inhib. Cone. (meg., ml )
Aerobacter aerogenes 50 Staphylococcus aureus

Klebsiella pneumoniae 12 5 Proteus sp.

Escherichia coli 156 Pseudomonas sp.

Salmonella lyphosa 78 Brucella bronchisepticae

S. paratyphi 0-78 Mycobacterium ranae

The new products of the process of this invention have.utihty as antimicrobials
isms from pharmaceutical equipment or to separat certain species from mixtures of various types of microorganisms— In the treat
ment ot animals infected with certain organisms sensitive to the action ot the materials, viz, such gram- positive microorganisms.

19
50
12 -5
0-39
1 9

Streptococcus faecalis 19
BY

-In sterilization to remove susceptible microorgan

Figure 6

139

and Douglas (Figs. 1-4) as the complete ab-
stract of each patent, without requiring any ab-
stractor's services.

We will go into the mechanics of making these
drawings, and disclose what we have found to be
the best time-saving approach and apparatus.

The packet also includes a copy of the May 1 949
Journal of the Patent Office Society, containing
many drawings, a few of which will be referred to,
for example, those corresponding to the four 30"
by 40" blow-ups on the easel. Likewise, a May
1955 reprint of another J.P.O.S. article is present
(the post-patenting drawing on the tetracycline
patent). (Figs. 5 and 6.)

But, before going into this material in more de-
tail, we might stop to inquire a little more about the
title of this discussion. Particularly, we must now
touch on the "encoding and post-retrieval benefits."

The approach can be on the basis of assuming
that a suitable electronic or other machine and sys-
tem for retrieval of information is settled upon.
We could ask: "What kind of document will it re-
trive, where patents are the subject matter?" Will
patents as now generally published be a satisfactory
form as the ultimate goal of the retrieval process, or
should an effort be made to streamline the informa-
tion display of patents, to facilitate comprehension
of the art contents of the patent document after it
has been retrieved? The present discussion is based
on the view that a considerable amount of facilita-
tion of comprehension of the subject matter of
patents can be accomplished. In the chemical
patents, ( see any one of the Webber, Silver, Rogers
or Douglas patent drawings and the tetracycline
drawings, for example) the drawings of the form
here shown facilitate comprehension of the subject
matter by presenting the overall process and the
specific details of each step in the process, these
details being located at the particular step of the
process to which they pertain.

These patents were selected for your packets as
demonstrating in part, what can be done to facili-
tate comprehension of chemical patents, and quick
recognition of the disclosure of these patents. This
form of presentation is at present found in relatively
few patents, although it was first urged in 1948.

Therefore, I am assuming that, for post-retrieval
benefits, it would be desirable that patent docu-
ments be improved to facilitate comprehension by
the ultimate user.

But, any retrieval system of the kind being con-
templated requires encoding of the information to

be stored for retrieval. The encoding process in-
volves as an essential element, analysis of the docu-
ment and comprehension is the same at encoding as
it is at post-retrieval. Additionally, at encoding, it
is essential that all the art teachings of the patent
document be available for encoding. The encod-
ing stage is therefore the critical point, before which
the art contents of the patent must be in readily
available form.

Mr. Pfeffer of the organizing committee of this
workshop has informed me that the Research and
Development group of the Patent Office plans to
proceed to encoding of more complex information
than it has attempted to date, and that it will be-
come essential to convert the information of the
chemical patents they will be working on, to a more
readily comprehensible form. In his opinion, it
will be necessary to first put the information in some
form, such as that in the J.P.O.S. articles, which is
the form followed by, for example, Webber, Silver,
etc. It is contemplated that this will be carried out
in one of the projects in R & D. Mr. Pfeffer stated
that after this conversion is made, the conversion
product could be used for information exchange on
an international basis between patent offices of the
various countries.

Accordingly, we can also conclude that for encod-
ing benefits, it would be desirable that patent docu-
ments be improved to facilitate comprehension by
the encoder. This improved form should be pre-
served to obtain the same benefits post-retrieval.

At this point it would seem appropriate to look
at the blow-ups on the easel. They correspond to
pages 344, and 347, 366 and 367 of the May 1949
J.P.O.S. in your hands. This constitutes a striking
demonstration of the differences in comprehension
contributed by change in form alone to the straight
line vertical flow, right hand horizontal attack
form. Exactly the same information is contained
in both of these diagrams. It is clear that the in-
formation display or presentation on page 347
(Fig. 8), is tremendously superior to that of page
344. (Fig. 7.)

But the diagram of Fig. 8 would not provide
sufficient analysis for encoding, since all the details
of art are required to be available for encoding.
Likewise, for post-retrieval purposes, early ref-
erence to and time-consuming analysis of the written
disclosure would be required, since the metal in-
volved is shown only generically and the details are
few.

140

Now by turning to pages 366 and 367 (Figs. 9
and 10, respectively) we find in two sheets of draw-
ings essentially the entire disclosure of the patent.
It is clear that these two sheets would provide an
analysis of the patent to facilitate comprehension
for encoding and for post-retrieval use. In the
latter, or "ultimate consumer's use," the saving in
time will be effected on each and every subsequent
use, whether this use will be by patent examiners,
patent searchers, chemical researchers, attorneys,
judges, or anyone else.

At this point I would like to ask you to visualize
with me the following: Picture a chemical patent
examiner at work today. He has just picked up
an amended application and is preparing to give it
a responsive action. There is no drawing in the
application. The first office action has cited eight
references. The examiner might or might not have
been the one who made the first action. Let us
assume the best situation — he was the examiner
who gave the first examination. But that was a
year ago (assuming the examiner is acting on
amended cases six months after receipt ) , and he
has no strong recollection of the case. He reads
his first action. It makes good sense to him. He
then reads the attorney's response. Facts from the
eight references and arguments concerning them
have been assembled by the attorney and a con-
clusion reached by the attorney, directly contrary
to the examiner's in his first action. Now the ex-
aminer is no longer certain of his facts or grounds
of rejection. He must restudy the application's
specification and each of the eight references. This
can be a very great time consumer, but must be
done before the examiner can come to a valid
conclusion.

Let us change the situation a little. Assume that
the application as filed had a drawing "informa-
tion display" presenting the full disclosure along the
lines of the drawings at pages 366 and 367 of the
J.P.O.S. as referred to above (i.e., Figs. 9 and 10)
or like the tetracycline patent drawings ( Figs. 5 and
6). Further assume that each of eight references
had, similarly, a drawing or drawings presenting in
full the art disclosure of its patent. Wouldn't this
examiner be in very good shape to expeditiously
determine the facts? Extend this to all chemical
examination — all applications being filed with such
drawings, and all references having such drawings.
Do we have here a key to overcoming the "back-
log" — that is, to reducing the number of active ap-
plications present in the Patent Office — now about

200,000 cases resulting in extended average pend-
ency for all applications?

There is at present, as there should be, a great
foment of self-examination in attempting to find
an answer to the "backlog." It is clear that some
curative ingredient must be invoked. Minor
palliatives won't do. The question I would like to
see you ponder is whether the picture just invoked,
of applications filed with full information display
for almost immediate comprehension, and of patent
references, each with full information display for
immediate or almost immediate comprehension,
would help to take care of the "backlog," or in any
event, permit the Patent Office to truly carry out its
examination job more expeditiously? Extend the
picture to include mechanical and electrical patents.
In other words, picture a situation where each ap-
plication file contains its information displayed in a
manner for immediate or almost immediate com-
prehension, and each reference has its information
displayed similarly. Could not this be a good part
of the missing curative ingredients?

While picturing this, let us not lose the signifi-
cance of the above assumption of applications as
filed containing information displays for almost im-
mediate comprehension. In the United States
there are about 7,500 registered patent attorneys
and agents (and their mechanical drafting staffs)
and 83,000 or more applicants for patents per
year. If they can be trained to file with their ap-
plications information displays presenting their art
teachings for immediate or almost immediate com-
prehension, the Patent Office examining corps has
been augmented from its present 1 ,000 or so exam-
iners, by an external staff of 90,000 or so, while
reducing the time the examiner must spend on each
case he acts on, because of the improved informa-
tion display in applications and references. In
other words, the Patent Office will have succeeded
in "farming out" a significant part of its work to a
vast working staff throughout the country and in
fact, throughout the world. And it would be a
proper placing of a burden upon those seeking a
privilege, provided that the Patent Office does its
proper share by supplying a plan and instructions
for such information displays.

Can we properly feel that this is what we are
setting about doing here and now at this Workshop?

In expediting the examination work of the
Patent Offices, the importance of the role of
mechanical or electronic information retrieval can-
not be forgotten. Work must continue on it. But

141

June 26, 1945.

M. P. MATUSZAK

PREPARATION OF CATALYST AND METHOD FOR

DEHYDROGENATING HYDROCARBONS WITH SAME

Filed Feb. 5, 1942

2,379,172

ADDITION, WITH STIRRING. OF CONCENTRATED
AQUEOUS AMMONIUM HYDROXIDE TO
CONCENTRATED AQUEOUS SOLUTION OF
SALT OF METAL

SOL OF METAL HYDROXIDE

j D-ILUTION WITH WATER

'£

ADDITION OF SOL TO

DILUTE AQUEOUS AMMONIUM

HYDROXIDE NOT OVER O.I NORMAL

GELATINOUS HYDROXIDE

SEPARATION FROM SOLUTION

I WASHING |

DRYING

METAL OXIDE CATALYST

INVENTOR

MARYAN P. MATUSZAK

ATTORNEY

Figure 7

142

PREPARATION OF CATALYST AND METHOD FOR

DEHYDROGENATING HYDROCARBONS WITH SAME

Pat. No. 2,379,172

SALTOF METAL

Concentrated
Solution

add while

stirring

NH4OH

Concentrated
Solution

METAL HYDROXIDE
(sol)

to dilute

WATER

optional step

sol added

;o this scln

Dilute Solution
(not over O.lNormal)

GELATINOUS HYDROXIDE
(aq.)

Liquor *

off

Wash

Dry

optional Btep

METAL OXIDE
CATALYST

INVENTOR.

Figure 8

143

2. Sheef-5
Skeetl

PREPARATION OP CATALYST AND METHOD POR

DEHYDROGENATING HYDROCARBONS WITH SAME

Pat. No. 2, .'579,172

CN.CN.
O O

2 3

O U

-30) +i x) a

IE .C C cu CU

24J ftfi ft

•I

4> 4) S

•D Art

o
U I o

o C c»-

VlfiT) I

"o-h a o

CD P> UN,

w eg <m

3 C O P
CD 0)

CD bi) &0

X X -P X> o c!
O O rcJ ti -rl

s fc> °

a* n

f-,

0.1

<D S-.

bl)

O
1 -H

(1)

p P C\i

< IE CNi

•H

• In •

-w

-P o

alys
omic
2/1/

<l>

n nt<

'-^--* ni TJ

•rl

<^\ r^vr-l -

O O O C

S S H O

t-H

w*~- .H C

S-i i-l =

o < cr o

ti >

i-H

O r-H G

-P

O o W

CM -P 10 O

EC H 10

On 0) -H C

• a -o -H

CNJ
32

>
3

'"^

W; t. OH

g^>«,

•a

•rl

•H

P M

s 1

w G

•H

•rl
X

■H

S-i

CO CO

-3

■w

•H

e i-i

111

•rl CO

P r-1

-P C

(0 X

(-1

<H CO

n)'

m cr*— -
co cu u
=J o

riOll

a) r~- o

a
cu

"fl

ctl

S>>

tp (0

r^ O

P JO

•rl tl

> IT)

■rl O

i-l

■P o

*>>

JC J3

•H N-,

X O

i-
3

144

Sheet 2

PREPARATION OP CATALYST AND METHOD FOR

DEHYDROGENATING HYDROCARBONS WITH SAME

Pat. No. 2,379,172

I cd p a>

K x! 01 fl

O +5 rS ri

m 01

O XI

XI
0)

hydr
hen
atal
adso

CD XI

•rH t«
CD -P

w <3

S s w

oi rt X)

(15

-P

■rl

■d

rH <M

-p

•» m

ti X) (0 O

P CD X!

rH X)

•P c

(n a

f-i in

i -'

• H b>_:

CD -H O

rid

-P

P iH

O XI

0) in p oi

rj ifl

3 CD

CO P. O

3 -H CD

'rt -ri
en o

S3 -P

•n

CD
fl

■rl

•» to
en i

a e

•rl 01 -rt CO

C rH XI
•rl (0 -H

•a -p o

O 01

(Ti

1) r-\

'tO CD rH

r-\ U X)

a r".

rl JJ >

fl Q. C

•H

rH t.

to rt

X >* O H

o a <o

rr,

■a

•rt 01 3

4) M

3 § ss

•rt

H
o

P -rt CD

CD H P

nf o

oi in

m to to

IH (0 03

rHrH

>» CD 0)

9 % h

<T!

C XI U =H

^ .H a o

(!)

cfl

a. rt <h

!>»p

cn!

1-1

111

in! c\J rJ

CM01

■> CD

c-\r<-\ m

prV>

O O

o o rt

IS, s

2 3rl

£i££<H

x: c tn

E-i ro > N> csi

3 rt M
O U, O tH a

•p in
CD

C -H <H

PN3 O

XI
C

a 'rt

. p
in nl

xi -a

CD >»
P -^

to rt

"3 o

rf\ rr) C\J nj

"■»"" ^'~ » P

rr> rr\ P"\rH

01
(*> pry mp t\l CNl CM

CNfNrr-\ti r>r<^n
O O O xf O O O tryj o O

B 2S JS >>£: 13 S rH !3 Z 55

(daiO301<Urtbi>

"" Pn fa r-3 S

H 01

01 rH

n) to

XI P>

§p

C
«H CD

rH fn 01

<;o[i.

o cj cj

rH H

tU >
-P

rt •
°S

0) «H

no o

1 01

01 M

CM Q CD

ac n x:
-p &,
rt

•H CD

Q P

-P c
0) -H

•

rH 01
tO X)

-p «

to E

CD
H

•H

5
C 01

•rl CD

X 0)

rt -H

CD P.

•H

to O

gel cata!
' 20-li0 me:

Jieat grad-

rr\ p CJ

NrH O

HrllA
rH P U\

rt c

,s 3

0) 01

XI XI

■flTJ

o o

H h

CD
XI
•H

0)
XI

u
o

-p

" O
P

to
o

in O

P CO CD

!>> -3 Xi

X) O 01

<m a

E-i

< s

• 01

H &
o c

J- °

. o

HO p

<1 p

o
o

O 1A O

rH CA CJ

<c;

f-1

•^3

rH CJ

-P -iJ

CO to
CJ !>>

O O

«H

01 C

01 o

3 IH

c o

•pr

lA O

o o

-3 ho

u *

O Q

«H

xl oi

o >>

M "rt

■P

CJ (0

£.3

r-t

tO XJ

P CD

«} Jh

o rt
ft

•u

0) < s

P XI

-P -P MJ rH

oi oi

01 O It

rt 01

C f>> • O

tl)

•H rH O rH

P tM

tO CO

10 tn

01 p 01

-p

C 01

X) CO x: XI

rt cj p oi

M-H

£S

CO -P XI

•H rH

•

•H

pa nj rt to

fn O

CD 01

^ CO

C Si

■P >>-P 01

•H

CD CD

to rt -H

S fE -P

•p X)

r-\ >

CO CD

t»

>> C

rH o,ac

-P

P O

tO 0) CD O

x; 3

3 O

P E O —

P H

rt rt x ac

CJ CO CD S

•g^

■H
>

CO (3

•H x!

• >

a cj

P Op
<^ OU\

0-3

rH CM -3

t-.j

<
E-i

E3
i

rH<

P P

rt co

•H

Ci !-->

cr;

-H

(11

•H

[fl

•H

(1)

,-Tl

( )

■H

3
00

o o

CMCM

x ac

ON ON --^
• • o*
er\ t^al

o o

3 ^i

X! XI 3

■P O

•H bOXl

r-i

o u

T) 'u c$

O CD

•p

01 IH

to rt

O XI -H M

U r=

o xl -p

rt 01 -rl

-3

•rt

1-1

H P CJ

01 O

0) o

CtOcO
z) J.

3 N.
10 liters
(sol added
slowly to
this NH^OH;

0)
rH

•r-l

t.H

CH
O

o
q
&

•H
01

•ri

•P M

01 o.

01 >

oi a>

XI 'O

X X
o o

rH U

•

<H
O

r '

fil

>•

super-
natant
liquor

f-1

O
=1
a-
■rl

//Vl

CD CO

X rH

nl rH

O rt

-P

U as

145

information display of the sort discussed above is an
immediate need, and will facilitate encoding for
such mechanical or electronic information retrieval,
while it is being worked upon and when it has been
perfected.

The actual path by which information display, of
the kind suggested, will come about is more likely
in the reverse order — that is, R & D will use it to
facilitate its encoding and then make its results
available. In so doing this, a body of experience
will be accumulated as guidance in directing the
Patent Office's "external staff" (referred to above)
in how to execute the information displays.

Abstracts

We mentioned at the start that a suggestion will
be made that the chemical abstract services could
use the drawings of patents such as those of Webber,
Silver, Rogers and Douglas, and of the tetracycline
patent, as the complete abstract of each patent,
without requiring any abstractor's service.

As we spread these drawings out before us we
can see that they are in fact abstracts, but a special
kind of abstracts. They are selective intensity ab-
stracts, if you will. Consider the tetracycline draw-
ing (sheet one). The non-initiate might object
that the drawing "shows too much" and he would
"just as soon read the specification." The answer
of course, is that when he reads the specification,
the material he reads will mean nothing until he
relates each part of it to the particular portion of
the multi-step process involved. But the drawing
has already done this and has set such material next
to the "primary feature" to which it relates. The
"selective intensity" terminology arises from the fact
that the one viewing the drawing immediately
directs his attention to the heavy skeleton outline
displaying starting materials, attacking chemicals
in the sequence applied, final product and inter-
mediate products. These have been called the
"primary features" and are the first rank of the
intensity of abstract. From these, the viewer can
determine either that he is not interested any further
in this document, in which case he need look no
deeper into it, or that he is interested further in the
document. In the latter case, he can view the
document in its second rank of intensity of abstract-
ing, by considering the details of information
grouped about each of the primary features to
which they relate, what we have previously termed
the "secondary features." In other words, the ab-

stract is made in such form that the reader selects
the intensity with which he wishes to pursue the
abstract, having a positive guidance device (the
primary features) which directs him to drop the
document at this stage or to pursue it further.

The Official Gazette of the U.S. Patent Office
provides a preview of what these drawings would
look like if used in the abstract services. I have a
number of pages from the Official Gazette, having
these drawings on a reduced scale.

I also have here for inspection an abstract of this
sort which I made in 1955. It is fitted into a
column of Chemical Abstracts, over the single para-
graph abstract, and contains the same information,
but in the display form herein advocated. Al-
though it is slightly longer than the solid paragraph
under it, it is essentially the same size and could
have been crowded a little vertically. This sample
was made up to provide a direct comparison in form.

On Claims

I have another item on information display which
I will circulate. Here it is. Interest in it is revived
by the recent notice in the Official Gazette (770
O.G. 417; September 19, 1961) suggesting the use
and printing of claims in tabular form. This item
was also made in 1955. It is a proposed publica-
tion, and is concerned with putting chemical process
claims in the advocated single standard form of
display. I have personally done much with claims
in this form, and have found such use extremely
beneficial in examination of patents. One of the
contemplated advantages for this publication was
the acquainting of patent people with the advan-
tages of casting chemical process claims in this form.
One reason for not carrying this publication through
to regular use was that it would be doomed to
failure if the idea of putting the entire disclosure in
such form were adopted, since the Official Gazette
would then carry such diagram of the entire dis-
closure, and there would be no need for Official
Gazette diagrams of claims. This little gem is en-
titled the "Chemical Process Companion (Unoffi-
cial ) , to Official Gazette of the United States Patent
Office," and is jacketed in the pretty blue paper of
the O.G. Three pages from the O.G. Companion
showing two possible ways to present the information
are shown (Figs. 11, 12 and 13).

Extension of Principles to Mechanical Patents

In the early remarks we stated that there is
material in your packets on mechanical patents,

146

March 15, 1955

CHEMICAL

363

2,704,239

PRODUCTION OF SODIUM BICARBONATE AND

SODA ASH FROM TRONA

SODIUM CARBONATE
SODIUM BICARBONATE
(oq.)

iocfe

return to
system for
re-cycling

Mother
Liquor

unsaturated mother sol'n
NoBicorb. . 46 to . jfi.
Na Carb. "20 ~I6
by weight at20°CIOCPC

.dissolve

add dur-

ing the
cycle

CRUDE TRONAl

in amount substantially
equal to amount neces-
sary to convert the
Na carbonate of
the trona dissolved
during each cycle

into Na bicarbonate
and to maintain the recyc-
ling aq. sol'n corrfg. the
dissolved trona in a
conc'n within the system

Na 2 C0 3 NoHCOaH20
at which Na bicarbonate
is the stable crystal
phase in the surround-
ing liquor

crystallize
separate f

SODIUM BICARBONATE crystallized

2,704,240
PROCESS OF PREPARING TUNGSTIC OXIDE
FROM SINTERED MASSES CONTAINING
TUNGSTEN CARBIDE

ITUNGSTEN CARBIDE [ sintered mass

Heat
I300°F to
20OO°F

to cause

mass
disintegrate

^OXIDIZING ATMOSPHER"e]

TUNGSTIC OXIDE
METALLIC IMPURITIES

Metallic
Impurities

loose mass of
friable materials

- ANORGANIC ACID I

ITUNGSTICOXIDEI

2,704,243
METHOD OF DEFOLIATION

COTTON PLANT| or other plant hawing foliage

qpp'y

2tpl0
Ibs./ocre

SODIUM CHLORATE

SODIUM POLYBORATE

(aq. solution)

25 to 60
75 to 40

DEFOLIATED
COTTON PLANT

NogOXB^),

or other corres- x
ponding plant

-P

4. 16 to 7

2,704,247
METHOD OF MAKING LOW CARBON STEEL

I STEEL- MAKING INGREDIENTS
ISLAG-PRQDUCING INGREDIENTS

Cha rge,
containing

melt
down
charge

Entire process carried out in

ELECTRIC FURNACE
Powered by carbon electrodes

IRESULTING MELT

ref inej to materially reduce the

corbon content of the melt

IREFINED MELTl

power off

IDESLAGGED MELTl

in furnace

deliver

to melt
to cover

METALLIC
ALUMINUM

SLAG-PRODUCING
MATERIAL

melt

to blow

" melt,

to
completely oxidize the

metallic aluminum

OXYGEN
(gaseous)

power on| (full), in furnace

, ,Fully reduce the slag on the melt

|LOW CARBON STEELl

2,704,248

METHOD OF SEPARATING FERROUS METAL

FROM ITS GANGUE

IRON ORE
(solid)

Refractory

lined

furnace

gas furnace

with heating

tuyeres

direct

from the

REDUCING FLAME

same

furnace and

charge

tuyeres against the Iron ore

to reduce and partially melt the ore.

■ Cause the f tame to become oxidizing:

frflfl HOXIDIZING FLAME]

SdlTIA ' '

charge

same
tuyeres

to partially reoxidize the iron and
produce additional heat within
the furnace

to aid In completely
melting the charge.

< Make the flame reducing again:

furnaceondo-l^^- lREDUCING FLAME]

tuyeres

to complete the reduction of the
partially oxidized molten iron.

IRON ORE
(molten)

the heated and
melted product.

Figure 1 1

147

March 15, 1955

CHEMICAL

365

2,704,249

METHOD FOR SEPARATING COMPOSITE

ALUMINUM-IRON ARTICLES

Com-
posite
articles

ALUMINUM
Bonded to

IRON
J

or aluminum alloys
or iron alloys

Heat the temperature and time of heating
at about being sufficient to cause form-
I050°F ation of a brittle iron aluminum
3 Hours I alloy layer

1 only at the interface

ALUMINUM
| Brittle-Bonded
IRON

— I

tumble| the articles,

to fracture the interface
thus causing the alum-
inum to become de-
tached from the iron

2,704,252
PHOTOMECHANICAL PROCESSES

PHOTOGRAPHIC ELEMENT

having a
HIGH CONTRAST EMULSION

expose element

to light from

subject

(^fg flOR.G.NAL SUBJECT!

i screen

[SCREEN LATEI
IMAGE

Development
such as to pro-
duce a dye im-
age with a
gamma value
of at least approx
imately 2.5

tp develop

screen
latent
image

and balance of element

AROMATIC PRIMARY]
AMINO DEVELOPING
AGENT
in presence of a
COLOR COUPLER

The color coupler com-
bines with the oxidation
products of the developer to
produce a dye image.

NEGATIVE SCREEN IMAGE ... - .

(silver image plus a balanceof element
dye image)

INEGATIVE SCREEN IMAGE
| (dye Ima ge, only)

made up of areas of developed silver
with a dye image in situ therewith

bleochi _, _ . -,, i
sliv ^ BLEACH l
Image

and balance of element

The light is in-
cident on same
side of photogr-
aphic element
as original ex
posure

Development to
a gamma of at
least approxim
ately 4.0

to expose

he unex-

Silver fialide re-
maining below
the negative
image

Any silver hdidc

rerrwin*abek>f jr^

the positive 1*1-3

Image
Any re-halogen -

bed silver of

the negative

image

of a color to which
LIGHT! * ne photographic
emulsion is sen-
sitive and
of such intensity
that the light is
substantially com-
pletely absorbed by the dye
image ft its maximum density.

posed areas of

the photographic

element

EVELOPERlSuchos will
Tfol redevelop

any rthalogenized

silver negative image

which may be present.

— which halide has been optically
screened from the second
exposure.

to remove

Negative

Dye
Image

re-

move

SCREEN POSITIVEI

Figure 12

148

March 15, 1955

CHEMICAL

363

2,704,239

PRODUCTION OF SODIUM BICARBONATE AND

SODA ASH FROM TRONA

Robert D. Pike, Greenwich, Conn.

Application June 1, 1951, Serial No. 229,437

10 Claims. (CI. 23—64)

1. A recycling system for the production of sodium
bicarbonate from crude trona by crystallization of sodium
bicarbonate from an aqueous solution of trona which
comprises

unsaturated mother sol'n

2,704,243

METHOD OF DEFOLIATION

Frank J Seibert, Bound Brook, N. J., assignor to Chin-

man Chem.cal Company, Inc., Bound Brook, N J a

corporation of New York '

No Drawing. Application April 10, 1952,

Serial No. 281,662

2 Claims. (CI. 71 — 2.2)

which A co n mp h r Ss° f def ° ,ia,, ' n8 C °' t0n a " d ° ,her P Iants

[COTTON PL A NT I or other plant having f olloge

'o
25 to 60

75 to 40

SODIUM CARBONATE

SODIUM BICARBONATE

(oq.)

I0CTC

return to
system for
re-cycling

Naj3icarb__4J2.to.i8

NaCarb. "20 16
by weight at20°CIOCPC

„ °pp'y

2 to 10
lbs. /acre

DEFOLIATED
COTTON PLANT

SODIUM CHLORATE
SODIUM POLYBORATE 75
(oq. solution) ^k

NagO.CBgOgV

or other corres- *
ponding plant

4.16 to 7

.dissolve

odd dur-

ing the
cycle

CRUDE TRONA]

in amount substantially
equal to amount neces-
sary to convert the
Na carbonate of
the trona dissolved
during each cycle

into Na bicarbonate
and to maintain the recyc-
ling aq. sol'n contg the
dissolved trona in a
conc'n. within the system

Na 2 C0 3 N0HCO3H2O
at which Na bicarbonate
is the stable crystal
phase in the surround-
ing liquor

2,704,247
METHOD OF MAKING LOW CARBON STEEL
William G. Connor, Tarentum, Pa., assignor to Allegheny
Ludlum Steel Corporation, Pittsburgh, Pa., a corpora-
tion of Pennsylvania

No Drawing. Application November 13, 1950,
Serial No. 195,489
7 Claims. (CI. 75—12)
1. A method of making low carbon sfeel, which con-
sists in

Mother
Liquor *

crystallize I
separate I

[SI

|sL

STEEL- MAKING INGREDIENTS
[SLAG- PRODUCING INGREDIENTS

SODIUM BICARBONATE crystallized

2,704,240
PROCESS OF PREPARING TUNGSTIC OXIDE
FROM SINTERED MASSES CONTAINING
TUNGSTEN CARBIDE

Howard S. Avery, Mahwah, N. J., assignor to American
Brake Shoe Company, New York, N. Y., a corporation
of Delaware
Application October 23, 1951, Serial No. 252,760
5 Claims. (CI. 23—140)
4. A process of recovering tungsten and metallic im-
purities from a sintered mass of tungsten carbide com-
prising the steps of

ITUNGSTEN CARBIDE W intered mass

Charge,
containing

Entire process carried out In

ELECTRIC FURNACE
Powered by carbon electrodes

JRESULTING MELT

refinel to materially reduce the

corbon content of the melt

IREFINEDMELT

|siog> —

power off I in furnace

IDESLAGGEDMELTI

Heat

I300°F to ^g ^OXIDlZINGATMOSPHEREl
2000°F disintegrate

TUNGSTIC OXIDE
METALLIC IMPURITIES

Metallic
Impurities

loose mass of
friable materials

ANORGANIC ACID]

.deliver

to melt

METALLIC
ALUMINUM

.to cover

* melt
to blow

SLAG -PRODUCING
MATERIAL

OXYGEN
(gaseous)

melt

completely oxidize the
metallic aluminum

joweronl (full), in furnace

TUNGSTIC OXIDE

Fully reduce the slag on the melt
|LOW CARBON STEELl

Figure 1 3

149

March 15, 1949

H. E. VEIT

METHOD OF FORMING RIVETED JOINTS
Filed June 8, 1945

2,464,622

In joining sheet metals, particularly in the aircraft industry,
it is customarily practice to dimple depress all rivet holes

so as to allow for fitting of the rivet heads flush with the
surface of the outside sheet presenting a smooth surface to
the air stream.
It is a fairly simple matter torivet dimpled sheets of relatively

soft alloys such as those corn-
However, in riveting dimp-
led high strength alloy sheets
containing a high percentage
of magnesium or similar hard
metal alloys there is a ten-
dency for the sheets to
crack outwardly around the
rivet holes.

Apparently this failure of
the metal is due to the lack
of ductility of these hard alloy ';
metals and their inability to £j
readily conform to the shape **
of the expanded shank of the rivet

posed of a high percent of aluminum.

MAGNESIUM ALLOY SHEETS

HAVING NESTED APERTURED

CONICAL DIMPLES

[

The present invention provides
a method whereby a flat washer of
a given ductility is placed around a
rivet and in contact with the inner
sheet of a plurality of sheets being
riveted together so that the strength
of the washer is utilized to effect
a collar-like action around the rivet,
and absorb the excessive force
developed in the rivet shank
during riveting, expanding
gradually to the shape of a
cone, hugging the dimple and 1
uniformly distributing the
riveting load over a wide
area of the dimpled portion
of the sheet being riveted.

I have discovered that the expan-
sion of the washer outwardly from
the center aperture thereof effected
by the expansion of the rivet shank
acts to progressively cup the wash-
er to the shape and contour of the
dimple.

This action results in a
uniform gripping of the dimpled
areas of the sheets under com-
pression.

especially thin sheets such as 52-SH and par-
ticularly in thickness range .020 to .040 inch

to be riveted with standard 3/16 inch, and

larger diameter rivets.
Sheets of these thicknesses apparently have
insufficient strength (tensile) in the metal sur-
rounding the rivet hole to restrain the shank
of the rivet as it is expanded during the riv-
eting operation

with the result that cracks and fissures

are formed in the dimpled area adjacent

the rivet.

Position in

said apertures
with the shank
extended through
the apertures
and beyond the apex
of the bottom sheet
and the shaped head seated flush
with the outer surface of top sheet

Preferred is of currently stan-
dard annealed aluminum
alloy, known in the art as
56-S, and may be 3/16 inch
having a in diameter,

shaped head and a shank portion

ALUMINUM ALLOY
RIVET

Place on the

FLAT METALLIC
WASHER

The washer is of aluminum
alloy, annealed to the same
extent as the rivet, and has
an overall diameter less than
the maximum outside dimple
diameter.

This latter dimension is
important in that I have found
that the washer gives the best
Apertured, having results if less than the
size of the dimple area
so that when shaped to
its final position, the
outside periphery of the
washer clears the edge
of the dimple and does
not overlap onto the plane
surface of the inner sheet.

shank of the

rivet and in

contact with the

apex of the expos- substantially the same

ed conical surface flow characteristics as

of said dimple the metal of the rivet

This assures a un-
iform expansion of the shank
of the rivet and the washer when subjected
to force. Furthermore, use of parts hav-

ing the same metal alloys obviates the pos-
sibility of undesirable galvanic action be-
tween the rive t and washer. Also, use of aluminum alloy rivets with

Apply

Aligned force
to the shank

magnesium alloy in the combin-
sufficient to compress ation herein are known to have
and expand said shown high non-corrosive char-
shank to a pear- acteristics under standard cor-
shaped form rosion tests conducted by the

outwardly against the aperture of the United States Government,
washer

todifferentially expand said aperture
to change said washer from a flat

to a conical shape, and
to effect downward movement of
the washer against the conical
dimples

Apply

Further force
to the shank

RIVETED
MAGNESIUM ALLOY SHEETS

so as to form

a second rivet
head thereon,

having an rverlap of
less than one-half the
radial width of said
washer

Edges of dimpled rivet holes protected against cracking

and breaking when riveted together
Galvanic action diminished
High non-corrosive characteristics

150

Figure 14

Oct. 23, 1956

H. E B SCOTT

DOCTORS

2,767,529

Filed Dec. 14, 1954

»Bevel, provided on front end of the blade
Doctor blade-

Pressure plate*
(applies pressure to the blade 11)

Holder— i

Bracket — — ^
(Biased by a spring, or by in

gravity, to cause the blade to Jz
bear on the roll 13)

Fig. 1. SIDE ELEVATION

SHOWING THE DOCTOR
BLADE IN POSITION ON A
ROLL OF A PAPER MAKING
MACHINE

Laminations
of woven or felted
glass fibers
(or other frangible
laminations)

Leading edge

Slight depressions, formed by
the breaking up of the fran-
gible material (of 15)

Upper lamination^
of finely woven cotton fabric )

having about 100-120 threads .'_
per linear inch 1/

Laminations of coarsely woven fabric,
or other wear resisting laminations,
(may be of coarse weave, having
about 40-60 threads per linear Inch)

Heel of blade

The laminations are individually impregnated
with a synthetic resin, are bonded together to
produce a blade of desired thickness,
e.g., 1/8 or 1/32 inch

Lower lamination
of finely woven cotton fabric
having about 100-120 threads
per linear inch

Fig. 2. GREATLY ENLARGED CROSS-SECTION
THROUGH THE BLADE

tiara /J, £. ff. s ca ft;

" — ^ rrV.

Figure 1 5

645803 0-63-6

151

wherein we applied the principles learned in work-
ing with chemical patents, or serving as a basis for
the results arrived at in working with chemical
patents. As to the latter, we have in mind the
principle that the drawing in a patent case should
make it unnecessary to refer to the specification, or
delay as long as possible such necessity to refer to
the specification to understand the drawing. We
would set as the aim of the drawing in a machine
case, the principle that it should be so well marked
with legends that it is unnecessary to refer to the
specification at all to understand the drawing and
the invention. Just as in chemical patents we have
recommended the putting of all detailed informa-
tion relating to a given point of a process, at that
point of the overall process, so would we seek in a
mechanical process case to do the same, where the
process lends itself to such display. See the work
in your folder on the Veit patent which describes a
method of forming riveted joints (Fig. 14). The
other possibility is shown in the Scott patent on a
doctor roll (Fig. 15) where all the description of
the specification relating to number designations of
the parts shown in the drawing are to be placed right
on the drawing, as well as explanatory matter not
designated by part-number legends. In an appro-
priate case, a combination of both displays might be
called for.

Tentative conclusions, mainly on the basis of the
work on the Morton patent, but also on the basis
of the Scott patent and other patents, were reached
as follows:

Machine patent drawings display numerical
legends designating parts referred to in the body of
the text of the specification. The numbers so used
are required to be displayed so that they are free
and clear of structure and require a lead line to the
part designated. Being thus free and clear, they
are fair game for attaching another lead line to
terminate in the textual matter of the specification
referring to a designated part. This textual
matter should be put directly on the sheet of draw-
ing, at nearby clear spaces. In a more complicated
case, the new lead line and textual matter could be
in the form of a transparent overlay. Overlays in-
volve problems which are avoided by direct appli-
cation of information to the main sheet, so that the
latter would usually be preferred.

The textual matter would be clear of the machine
as drawn. Only the new lead lines from the textual
matter to the part number would run across some
part of the drawing for part numbers not outside

the drawing. Lead lines, as used up to now, have
been beneficial and necessary. The additional
lead lines and proposed textual legends would seem
to be in the same category.

Expressing this as a laboratory or workshop prob-
lem, we could phrase it as follows :

Object : To comprehend in the shortest time
possible the teachings of the patent disclosure, down
to the finest detail shown, particularly for machine
patents.

Method : Have as a permanent part of the draw-
ing, (or less preferred, as a readily available acces-
sory, as by a transparent overlay ) legends from the
specification, describing the parts designated by the
numerical legends and describing relationships of
operation.

Reason : Each reader, in one manner or another,
fixes the relationship of each of the components of
the drawing to the disclosure of the specification
relating thereto, preliminary to comprehension.
_ By placing this on the drawing as it issues, a
permanent saving in time is effected by being built
into the patents, the saving being effected each time
the patent is picked up for first reading, particularly,
and each time thereafter for each individual reader.

Broadening the Base

To lay the basis for broadening the treatment of
chemical process patents or other chemical process
documents, to include at least some mechanical
processes, we will first very briefly review our treat-
ment of the chemical documents. This is how it
was summarized in the tetracyline reprint on page
381:

"The main purpose was to engineer patents . . .
to enable the placing in accessible form of all the
details of art in each patent." The single standard
pattern therein disclosed and proven universal for
all chemical processes was selected as the means for
reaching the stated goals. This may be stated in
brief as involving the specially defined terms "pri-
mary features" and their position in the "master
pattern" or "standardized pattern," as shown by
Fig. 16a.

The "secondary details" are all the data other
than the "primary features," and include the alter-
natives, if any for the primary features, the physical
treatments, data for the particular chemical, such
as concentration, pH and temperature.

These secondary details are gleaned from the
patent disclosure and placed on the flow sheet draw-
ing at the primary feature to which they pertain.

152

[ STARTING MATERIALSl

- 1 1st CHEMICAL!
- 12nd CHEMICALI
■ 43rdCHEMICAI_l
- i4fh CHEMICAL!

[final product!
the primary features

Figure 16a

The prior articles also recognize the aid which
ntermediate condition identification could give, and
lote many of the prior drawings for this.

Having thus reviewed the past treatment, we
proceed to the broader treatment as follows :

The Module of Process Information

A chemical or mechanical process or method in-
/olves the change or transformation of a starting
material produced by subjecting the starting ma-
terial to the presence or action of another material
3r of a force. The process may involve change of
institution, condition, association, position or shape
af the starting material. In all of these, except
association in a mechanical sense (which might be
called "assembly" ) and change of position, a trans-
formation of the starting material is involved. The
transformation may be chemical, as in change of
constitution, or it may be physical, as in change of
condition (or state) or shape. Changes in associa-
tion may be regarded under some circumstances as
chemical transformation, as for instance in a forma-
tion of a complex or "association" compound.

The common thread discernible in a process,
whether it be chemical or mechanical, is the existence
of a starting material, a contacting or influencing
material or force, and an end product. This makes
it possible to diagram processes of such kinds by a
common pattern.

We shall call the diagram illustrating a single step
process a "module of process information," since
it may be regarded as the unit for analysis of any
process ( Fig. 1 6b ) . Every complex or multi-step
process may be reduced to a succession of such
modules, the end product of the first module being
the starting product of the second module, and so
on, until the final product of the multi-step process
is reached (Fig. 16c).

As in the previous treatment of chemical process
patents, we would place next to the "primary
features" of the modular process information, the
"secondary details", that is, we would place next
to each box of the module, all the details of the dis-
closure pertaining to such box. We have attempted
to do this for the Veit patents (Fig. 14).

Conclusion

Patents recast in the forms urged will infinitely
aid manual searching (which may always be with
us), do-it-yourself punchcard encoding (such as
McBee Keysort and Peek-A-Boo, which should be
encouraged ) encoding for the big job and compre-
hension of the contents of the patents.

In conclusion, I would like to say that we know
that there is concern for the preservation of the
patent systems as systems of examination, prior to
issuance of the patent. It is fitting to observe this
now, since we are celebrating the 125th Anniver-
sary of the setting up of the U.S. Patent Examina-
tion System. The sorry state into which patents
had fallen due to issuance in a registration system is
described in an article by Mr. Rosa entitled "Patent
Office Organization," J.P.O.S., June 1949, Vol.
XXXI, No. 6, pages 414-448.

If we make of patents, documents in which com-
prehension has been facilitated, we are helping to
preserve the patent examination system, since we
are helping examination both timewise and
qualitywise.

Dr. Davis: Thank you, Mr. Wolk. As I under-
stand it, you are really talking about a new literary
form. I am not sure it is going to be as popular as
"Ulysses" was, or some of the other experiments in
literary form, but it certainly is an interesting de-
velopment, and I am sure you want to discuss it.
Is there discussion of this paper?

Mr. Jancin: I think there is a great deal to be said
for this approach, and it probably could apply to
systems inventions in block diagram form. How-
ever, we found that many overseas patent offices
object to the use of legends, and I really do not know
why. So I would like to direct my question to this
board right here. I am talking of the use of
legends as are shown with the various blocks on
the patent sheet.

Mr. Simons: Could I say, speaking for the Cana-
dian Patent Office, we try to have blocks labeled
with not only the designation, but particularly the

153

The Module
of Process Information

STARTING MATERIAL

CONTACTING MATERIAL

}

• material or force influer
the starting material

END PRODUCT 1

Figure 16b

Module 1

Module 2 i

Module n

STARTING MATERIAL

■

CONTACTING MATERIAL 1 1

or etc.

END PRODUCT 1

CONTACTING MATERIAL 2

}

or etc.

END PRODUCT

•

CONTACTING MATERIAL n

or etc.

FINAL PRODUCT

Figure 16c

154

jrminology, so there would not be any trouble in
ur country.

)r. de Haan: It is not that we do not allow the use
f legends but we cannot oblige the attorneys to
se them under our rules. And in general we do
ot like to have non-uniformity in the practice,
^ith our new law we will have new rules which will
robably go in this direction.

(In response to questions from Mr. O'Day, Mr.
^olk stated that he could not estimate the time it
'ould take a trained analyst to prepare the informa-
on in a patent according to his scheme and that he
ad used this procedure on chemical information
ealing with large, variable structures such as
larkush formulations. )

h. Tate: It was suggested that this might be use-
il to services such as Chemical Abstracts to
mplify the process of getting the material into the
xondary journal more promptly. Two things in
lis regard: One is that there are two types of
^rvices involved. One is an alerting service for the
eneral public in which you want to get the infor-
tation across promptly. The other is a retrospec-
ve search tool with which chemical abstracts is
irgely concerned at the moment. This type of
ow diagram does not seem to suit either purpose
:>r our general public, first, because it's quite ex-
tensive to produce. If you are going to produce a
heet of this size, it is going to be very expensive to
et it out and get it out promptly. Second, it has
luch too much detail for most retrospective search-
lg, because you wish to get back to the original
ocument. You do not want to use the journal
self as a stored document. And the result is that
his would be a great help to us, but it would be

help in that we could use a lower degree of train-
ig in the person doing the abstracting, to give us

document which we could then use to publish.

Ir. W oik: You say you would use a person of a
jwer degree of training. The point is you do not
teed to use anyone of any training. You just need
. printer to print the diagram, which you can find
n the O.G., assuming that the O.G. prints them
nore legibly. And I cannot see why it would not
atisfy the alerting service. It gives you the starting
naterial, the end material, and what goes in.
What more do you need in alerting? As for a
etrospective search tool, it has all the information of
he patent. It has the concept of ranking of infor-
nation and of variable intensity abstracting. I do
lot see why it would not serve as a retrospective

search tool. I think it serves both purposes. And
as for the sheet size being too large, I intentionally
passed the O.G. around to give you an idea of the
size it would be. You can get about four of these
on one sheet. And if you saw the sample of the
Chemical Abstract that I did there, we have a dia-
gram about the size of the printed block paragraph.
Why is that too large? I could compress that a
little more and get it to the same size. I think it
is an unduly harsh test that something that presents
the information in better form, occupy exactly the
same or less space than a solid block print in the
same size of print.

Dr. Tate: I am not objecting to the amount of
space or the amount of information you give. I am
objecting to the amount of information which is
given for the purpose to which it is to be put. For
your purpose, it certainly serves admirably perhaps.
For ours, it does not. We do not want to have a
stored document printed as an abstract.

Mr. Rot kin: I would like to pursue that just one
step further. Why don't you want to have the
source document as the abstract if, in fact, it is in
such form that there is no need to abstract? I am
assuming that there is no need to abstract. But if
you have it, particularly if there is going to be any
use of microfilm technique, because of the volume
of material and the space involved, this would be
ideal.

Dr. Tate: You are speaking about something in the
future, and I am speaking about something that is
now present. We are not speaking of microfilm
techniques for the journal as it is now constituted.
We are speaking about what it is going to be in five
years, and this is a different question. I am not
prepared to speak for that.

(Mr. Freudenberg, President of the Patent Law-
yers Club, announced that Mr. Kalikow would
address the organization that evening on the sub-
ject of a long-range program for mechanical search-
ing and extended an invitation to the Workshop
participants to attend. )

Dr. Davis: Our next paper is by a person whom you
have met already, Harold Pfeffer. But I had better
tell you that he is a patent research specialist. He is
a chemist and an attorney. He has been with the
Patent Office since 1939 as examiner, classifier and
as a member of the staff of Research and Develop-
ment. He is going to talk about the problems in
the analysis of patents.

155

Mr. Pfeffer: Mr. Moderator, ladies and gentlemen,
Mr. Wolk was kind enough to refer to my encour-
aging remarks with respect to his work. This is
true. For those of us who have had the difficult
task of trying to analyze the information in a patent
document, we all realize the need to reform the
information into a concise type of format, such as
he has illustrated here. However, I hope that he
is going to forgive me for what I am about to say.
My talk today is not an offer of any suggestion for
international cooperation. But it is intended to
point out the pitfalls and the difficulties of analysis
as performed by human beings, and the great need
for greater use of machines in this area. I am
afraid that the remarks that I will make here will
stand as a sort of guardian angel with a flaming
sword or perhaps. as a demon with a pitchfork, in
front of the entrance to the hall of mechanized in-
formation retrieval. And in order to get in there,
you will have to get by these remarks.

In recent years, our investigations in mechanized
information storage and retrieval have centered
largely on organizing systems which are based upon
various search strategies. Comparatively little time
and effort has been expended on the input side of
the systems — namely, the analysis of the documents,
to determine their information content and to pre-
pare them for the subsequent step of extracting
some or all of the information content, prior to
transforming it into a machine acceptable form.
The state of the art, of the information retrieval
problem, is such, that we are in no position at
this time to belittle the importance of any area of
investigation. However, in the total picture, the
problems of analysis and extraction of information
content are far greater than those of systems design.
As far as systems design is concerned, we can always
determine the desired characteristics of the systems.
We can formulate an orderly procedure and subse-
quently determine what the necessary hardware
should be that could perform the necessary task.
If the hardware is not available, we can always de-
termine again whether existing devices can readily
be modified to give the desired result. If this is not
possible, we can always go back to the system itself
and make appropriate changes to fit the situation.
However, with respect to the problems of analysis,
we have an entirely different picture. Here we
are plunged into the elusive problems of human
frailties and the even more elusive problems of
human communication and how it is accomplished.
We constantly witness the difficulties of communica-

tion when people talk to each other face to face
There is always a need to establish a common basis
of understanding before ideas can be communicated
This is accomplished very often by the give and tak(
of conversation, gestures, facial expressions, and sc
on. How much more difficult our problem becomes
when the communicator puts his ideas into the cold
printed text of a document, and then disappear
from the scene, leaving other people to attempt t(
solve the puzzle of what he was trying to say, with
out being able to ask him to interpret his language

This report is based upon some general observa
tions of a group of analysts, who are organic chem
ists and who have been working in the Office o
Research and Development, and more specifically
based upon observations of a small group which wa
working on the HAYSTAQ project.

Since the state of development of the art is suci
that full reliance must be placed upon humai
capabilities for a significant analysis in depth, of th<
information content of documents, the question tha
arises first is, what kind of people do we need? I
becomes apparent almost at once, that the peopli
who are engaged in this type of activity must b<
skilled in the particular discipline or art with whicl
a selected group of documents deals. Whether th<
degree of analysis to be performed is as meager a
that required by a simple coordinate indexing sys
tern or as deep as that required by the most inten
sive searching operation, the major requirement o:
the analyst is the ability to recognize a concept, nc
matter how it is expressed. It is perfectly obvious
I think, that the inability to recognize a concept oi
the recognition of the wrong concept will result ir
a loss of information at the initial stage of the
operation.

As soon as we begin to talk about skilled people
we immediately come to the problem of cost. Th(
individual, who is charged with making the de^
cision at this point, is faced with a dilemma o;
choosing between skilled employees at a high cosi
or employees of lesser skill at a lower cost, with th<
attendant risk of losing information in the system
We must strike a balance somewhere. This, to 2
large degree, will depend upon the requirements oi
the system and its intended use. Ideally, of course
there is no question as to what the choice should be
Everybody strives for the greatest degree of per
fection which is permissible under his own circum
stances. But we have to face the realities of an;
situation. And the reality here is that cost become:
an extremely limiting factor. In our own experi

156

nee, we had to buy the skills that were available
o an extremely low budget. Where we would have
iked to have the skill of somebody of the equivalent
i a Ph. D., we had to take the skill available to
>eople who were recent college graduates.

An additional factor which complicates the
iroblem, of course, is the time factor. If one be-
gins an operation with employees of relatively low
kill, with the hopeful expectation of having the
svel of their skill raised by on-the-job training, then
here is a slow-down in the total productive effort,
yith a consequent increase in cost. In a crash pro-
gram, of course, such a situation would be un-
hinkable. In other situations the time limitations
re a little bit more flexible, but they still cannot be
gnored.

In a number of projects which we have had in
he Patent Office of the United States, the cost of
nalysis has been variously estimated as between $25
3 $35 per document. I would like to caution you
t this point, that these figures are based upon ex-
erimental systems. Even where the system has
een put into operation, it is still of a pilot nature,
nd is being used as a basis for further study. In
he HAYSTAQ project, where analysis in depth is
•erformed, in order to find all the chemical com-
>ounds which are disclosed in a document, the
verage time of analysis, based upon actual elapsed
/orking time, is about five hours. At a salary rate
f about $3.00 an hour, this comes to about $15 per
ocument. This, of course, does not take into ac-
ount the time for coffee breaks, other rest periods,
nd various delays that are due to the distractions
ommon to any operation where a considerable
lumber of people are involved, which, of course,
/ould raise the cost to a considerable extent. The
ost by the time we actually get the information into
he system, I would estimate now as being on the
rder of $80 a document, and these are not com-
itate figures either. We are, however, at this point
[uestioning whether the cost per document is a
iroper measurement since the size of the document
3 not necessarily proportional to the amount of
>ertinent information that it contains. A better
aeasure might be the cost per unit of information,
vhich, of course, will vary again depending upon
he determination of what that unit should be.
Jnfortunately, we do not as yet have figures of this
ype. So far as the Patent Office operation is con-
erned it is obvious that the cost can affect us to a
arge extent. We have heard a good deal this past
veek about the role of the Patent Office in the na-

tional economy. If the Patent system is to remain
a potent force in our economy, it must produce a
reliable product, and if the cost of that reliable
product is exorbitant, I do not see how management
can do otherwise than grit its teeth and accept the
cost, while it looks for something better.

Now, what about the people who actually do the
analysis? What kinds of personalities are required?
What restraints are imposed by the nature of the
work, on the selection of personnel? The work in-
volves a continuous concentrated reading. I can
tell you from my own experience that I would not
like to be doing this work for eight hours a day and
for five days a week. Under the best conditions,
this will lead to many physical and mental tensions
and strains. If you add to this the distractions of
various external stimuli, the situation becomes even
worse. In a period of about two years in the small
group of analysts that we have been able to observe,
we did from time to time witness mounting tensions
exploding into bitter arguments over trivial matters.
It also did not take very long before we found sev-
eral new sets of eyeglasses in the group.

The tension which is inherent in this type of work
results in the rapid approach of a state of fatigue.
In this state the effects of external stimuli become
even more pronounced and lead to a significant in-
crease in errors. I asked a number of the people
under my control to observe as closely as possible
when they began making errors which they them-
selves were able to recognize. The critical period
appeared to be after about two hours. Of course,
what we could not determine from this type of self-
examination, was the point where errors began
which were not recognizable by the individual doing
the observing. At any rate, in an attempt to over-
come the fatigue factor, the people were encour-
aged to take frequent short rests.

I also found that inadequate lighting was a con-
stant source of irritation, if I can judge from th<*
frequency and intensity of the complaints on thi;
point. The absence of a telephone on the desk ap-
peared to be another irritating factor, although I am
not sure that correcting this would not have resulted
in other problems which would have affected the
output. We thought at one time that a great many
of the disturbing influences could be eliminated by
providing privacy for the analysts. When this was
finally achieved, and each analyst was provided with
a room, I noted that there was a growing discom-
fort on their part, and that finally they began to
congregate while working. Still, in discussing the

157

possibility of sharing a room, each one showed re-
sentment at the idea. So it appears that some of
these causes of irritation are probably the result of
the de?ire of the individual to acquire some kind of
status symbol.

Now these observations are based on a more or
less casual, empirical study by an unqualified ob-
server. It appears, that in order to assure a smooth
running operation with a minimum of error in the
productive output, it will require some kind of
controlled, clinical experiments, to determine the
desirable personality traits and the physical char-
acteristics of people engaged in this type of work.
Experimentation will also be required to determine
the causes of fatigue and how to minimize them.

The large number of operations involved in han-
dling data from the time the documents are analyzed
up to the last transformation when they are finally
stored in whatever medium is selected, is conducive
to error-making. While some systems may tolerate
errors to a greater extent than others, it is always
desirable to reduce their occurrence as far as possi-
ble. In most of the operations involved, wherever
machines can replace human effort, we find that
errors can be minimized. However, the state of the
art requires that analysis of documents be performed
by human beings today, and probably in the fore-
seeable future.

In utilizing human effort, one of the ground rules
is that you must accept the fact that errors will be
made. The errors may be both of omission and
commission. They may result from a lack of knowl-
edge of the subject matter or from incorrect in-
terpretation of the text. They may result from
fatigue, and from the many diverse distractions to
which humans are susceptible. Well, what can we
do to minimize errors at this stage? Again I am
speaking from the position of one with a relatively
small amount of experience, but it would appear
that an important factor would be to increase the
motivation of the individual to put out a more reli-
able product, by making working conditions as
pleasant as possible, and by setting up some sort of
system of reward and recognition of the individual's
efforts. How far can we push such an approach?
Well, again I will ask how much error can we
tolerate in a particular system, and how much is
management willing to spend to reduce it?

Another means we have of controlling error is by
checking. Where more error can be tolerated, we
can use spot checking techniques. Where elimina-
tion of error is critical, we have to resort to the

unpleasant, slow and costly method of analyzing
each document twice, and independently, and thei
comparing the results and resolving the differences
This is the system we have used in the HAYSTA£
project, and in other Patent Office projects. I hav
always emphasized to my people and I know tha
the other project leaders have, too, that accurac
is always more desirable than speed in this type o
work, although the speed itself can not be complete!
overlooked either.

Now up to this point, I have talked about th
many factors which influence an analysis operation
Now let us look at the problems of analysis per se
The first determination to be made is the exten
of the analysis required by the system. This is ac
complished by deciding which set of concepts i
critical for the purpose of searching through a givei
set of documents in order to isolate some sub-grou]
of one or more. As the extent of the analysi
increases, the set of concepts to be recognized in
creases correspondingly. The problem of setting u
schedules of terms which will be recognition symbol
for these concepts, also becomes more difficult i:
direct ratio to the increase of the depth of analysis
The problem of choosing a proper set of terms is nc
as simple as it might appear at first. For examph
there is the possibility of a conflict between th
choice of an arbitrarily contrived concept, whic
would make life easier for the system designer an
one which reflects the experience of those who ar
skilled in a particular art. Our experience in th
past has been that if a great majority of the cor
cepts used is of the latter category, there is not s
much reluctance on the part of the user to accep
some of the former. This is something I think m
brother classifiers will recognize well, too.

It is also noted here that as the set of concept
increases in size, the sharp delineations among ther
begin to disappear. We can no longer deal wit
such things as black and white, but we must b
able to identify various shades of gray. As ovei
lapping appears, the problems of hierarchical class:
fication are imposed on the set. The difficulties c
attempting to organize a set of concepts to establis
their similarities and differences, is well known t
classifiers and I will not go into the matter here.

Now, having established the concepts which th
analysts are to recognize, the real problems c
analysis begin. The difficulties here are largely ir
volved in a correct interpretation of the author
communication, followed by a correct translatio
on the part of the analyst into a standard set c

158

ymbols which best define the expressed concepts.
^. number of these have been called to our attention
i our own investigation and I will mention them
iriefly.

One of these difficulties is the change in meaning
if certain terms over a period of time. In the
roup of patents that we dealt with, we had patents
ping back to 1840 and up to the present time,
'or example, the term anilido was found in some
f the old patents, and this term was used as repre-
ssing an aniline derivative. In the newer patents
nd in the present understanding of this term, it is
eferred to as an acid condensation product of
niline, so that the meaning is entirely different,
ind here, of course, we have to rely on the experi-
nce and the knowledge of the analyst and his
wareness of the time at which the document was
ublished, to arrive at the correct interpretation.

An analogous situation is the change in meaning
f a term when shifting from one art to another,
nd this, of course, was mentioned by Mr. Oredsson
liis morning with reference to the word spring,
low we have another example in the word ring,
/hich also may mean different things to various
eople. It has different significance to the organic
hemist, to the chemical engineer, to the automotive
ngineer, to the jeweler, to the sportsman. In
ther situations, the difficulty is caused by the fact
lat the document may be a translation from an-
ther language by one who is unfamiliar with the
iiom of the target language. The ambiguities in
ach a case can sometimes be resolved by a careful
erusal of the surrounding text. A correlary situa-
on is one where the same art term has different
leanings in the source language and in the target
inguage. For example, many years ago, the word
ether" was used by the French and Germans to
xpress a concept which was defined in English by
tie term "ester." In documents which were
ranslations from these languages, this term was
ery often not translated: since there is also a con-
ept defined by the term "ether" in English, the
nalyst had to examine the text very carefully to
letermine which of these concepts was really in-
ended. Now the peculiarity of this type of situa-
ion is that the analyst must recognize that there is,
i fact, an ambiguity before he can even begin to
esolve it.

Another area, which has been quite troublesome,
5 that of the latent or not specifically expressed
oncept. This has been found quite frequently in
he group of chemical documents with which we are

working. In describing a chemical synthesis for
example, the author sometimes describes the admix-
ing of a plurality of reactants, in a not too clearly
expressed time sequence, and then indicates the
resultant product. Now, in fact, the process may
actually be a plural stage process, with a recoverable
product which has been produced at each stage.
The analyst must, therefore, rely on his knowledge
and experience to determine that there is, in fact,
a multi-stage synthesis and must also determine what
the recoverable products are, even though they were
never mentioned by the author. This quite fre-
quently requires delving into source materials to
determine accurately the nature of the reaction and
what the products are.

There are also ambiguities resulting from typo-
graphical errors. Where these occur in the text, of
course, clues for correct interpretation may be found
in the surrounding text. Errors of this type may
also occur in illustrative drawings or in structural
formulas in the field of chemistry. Again it is
the job of the analyst to correctly interpret the
information.

Another problem is raised by documents where
information is incorporated by reference. For ex-
ample, we have found documents where the author
describes a particular process and then says "in
place of item "b", I may use any of the compounds
which are described in Patent No. so and so, or in
the article by X in the Chemische Berichte of the
year 1896," and so on. Such references to other
documents must be carefully examined to determine
whether there is, in fact, an incorporation of in-
formation or just a mere incidental referral. And
also to determine what subject matter specifically,
if any, is to be incorporated.

In some special types of operations, such as those
that one finds in the Patent Office, some of these
problems become even more aggravated. This is
particularly true with respect to the occurrence of
error. Because of the great need for reliable patent
grants, records which are prepared for Patent Office
searching must theoretically be 100 percent free of
error. An additional aggravation occurs in prob-
lems of interpretation. A Patent Office, after all,
presumes to speak with a voice of authority in its
decisions relative to the grant of a patent. It, there-
fore, should not be placed in the embarrassing posi-
tion of having to reverse an interpretation made by
its own employees. This is bad both for public rela-
tions and for its own prestige. So, where there are
apparent ambiguities, our analysts have been in-

159

doctrinated to make only such interpretations which
appear to be based upon reasonable inferences made
from considerations of the total text.

In conclusion, it appears that the problem of
analysis is quite critical and is complicated by a
large number of factors, which at the moment do
not lend themselves to solution by machine tech-
niques. The future of the poor analyst does not look
too bright, either. At best, there are the various
problems that confront him, as I have enumerated
them here. And at the worst, he can look forward to
being eventually replaced by some computer-like
device. I can only conclude here with the immortal
words of Gilbert and Sullivan, that an analyst's lot
is not a happy one. Thank you.

Dr. Davis: Any volunteers to be analysts as a result
of this?

Dr. De Haan: Mr. Moderator, I think we owe much
gratitude to Mr. Pfeffer for what he has brought
us this afternoon, for I feel, he has gone down to
the root of the question and confronted us with
facts. I would like to say that I do not speak on my
own authority. I have stated the conclusions to
which the classification division, studying automa-
tion in my office has come, which is almost exactly
the same conclusions to which Mr. Pfeffer has come.
So when two ways of thinking converge in such a
way, I think, there must be a good deal of correct-
ness in it.

I think it is important that Mr. Pfeffer has stressed
that the analysis of documents is the pith and mar-
row of the whole system and that our greatest worry
in arriving at an automatic search system is how to
bring the facts and the relations contained in a
document into the machine in such a way that we
can get an answer if we ask the machine if such
facts and such combinations are known. The
analysis of documents must be done with the utmost
care or machine searching to be a success. It has
been mentioned here, and I have heard it from other
researchers that they are trying to analyze docu-
ments automatically. Perhaps after having settled
some preliminary problems of mechanical search-
ing, some time in the future, neurologists will have
determined the exact working of our brains so that
utilizing their knowledge we may arrive at a more
systematic analysis of documents. But at the
present, documents are written by each man ac-
cording to his own character, and the characters
differ much and we do not understand them. I
quite agree with Mr. Pfeffer that it is impossible to

neglect in any way the problems of the analyzers c
documents, I think we can only start with what
might call manual analysis of documents, and nc
try to do it by the frequency of words which recu
regularly in a document. I think that is very uncei
tain if only the reason as printed out by Mr. Pfeffe
that words often change in their meaning.

I should like to draw the attention of Mr. Pfeffe
to his statement that checking is so difficult. Th
classification division in my office thinks that pei
haps the best way of checking the analysis of docv,
ments is to re-search applications already dispose
of to see whether the machine will produce a li;
of the same documents which were cited by th
manual searcher, you could deduce therefror
whether the analysis was not quite right .

Mr. Kalikow: I do not think I have ever heard tw
more eloquent arguments in favor of full text storag
of patents. I think the two presentations that yo
have just heard illustrate the tremendous problerr
that are involved whenever you begin to do an
kind of encoding or abstracting. I might say the
I spent nine months doing exactly this, encodin
2,000 patents, after which I came to the conclusio
that this would never really work. You must pi
in the full text of these patents, because you neve
know which way you want to get that informatio
out. There are many errors of omission and dii
ferences of interpretation that the abstractor bring;
and this is expensive. I think Mr. Pfeffer pointe
this out. But when you add up all the cost of wha
he is doing and you compare it against the cos
perhaps of putting the full text of a limited portio:
of our file into storage, I think they almost becom
comparable.

Mr. Pfeffer: I would like to answer Mr. Kaliko\
on this point. This is not really an argument, bis
merely a clarification. What we are doing is fc
an actual operation and what you are suggestin
is a further exploration into the idea of utilizing th
full text, because actually we do not know how t
do it as yet. This is an area for further researc.
that you offer.

Mr. Kalikow: This is precisely what I propos
ought to be done. (Mr. Frome suggested tha
graduate students used as part-time employee
would offer a partial solution to the problem c
fatigue as related to error making. )

Mr. Frome: We have the problem of analyzin
patents today. What Mr. Kalikow suggests ma

160

e the ultimate solution. But what are we going
> do for the next five or ten years until it is solved?

fr. Shera: I would like to take up that point that
as just been made. We have had some experi-
rce, not with patents, but with other types of ab-
racting work, using graduate students, we have
iund, just as you have, that not only is it a very
gnomical way of getting the work done, and a
;ry good way of getting it done, but also the chem-
ts and physicists tell us, it is an excellent form of
aining and education for the student himself.
I really wanted to address myself to another
Dint. I was a little surprised that Mr. Kalikow
ould say that here was the best argument for using
le complete text. It would seem to me it is just
cactly the opposite. It is the best argument
gainst using the complete text, because there is no
ich thing as a complete text. Someone has said
tat a book is very like a mirror. A jackass can
Dt look into it and expect to find a wise man look-
ig out. A document says nothing except what
e say it says. And I think using the complete
xt in a machine is going to miss completely, the
;ry point that you brought up in the changes of
leanings of words over a period of time. It does
ot make any difference whether "ester" changes
leaning whether it is in an abstract or in the com-
lete text. The problem is still there. Also be-
luse of the whole syntactical structure, and the
.tent element which you mentioned, which does
Dt come out in the text. So there is still, as I can
e it, no substitute for a re-working of the thing, so
lat these elements which you have mentioned come
at and are eliminated.

fr. Kalikow: I do not disagree that by using the
ill text you still have the problem. But there is a
emendous advantage in utilizing the full text in
lat the problems then become problems only at
ne area, namely, that of retrieval. Instead of
aving problems in the classification or the encod-
ig area, and then also having problems in the re-
ieval area, you then have transferred all of your
roblems into one area, namely, the retrieval area.
,nd you do not have the problem of trying to have
Dnsistency between the encoder and the retriever.

fr. Pfeffer: I would like to take issue with Mr.
..alikow on the problem of costs. On the matter
f putting the full text into a machine, I think the
ost at the moment may be even higher than that
f using a human analyst.

Mr. Kalikow: Oh I am sure it is. But I think it Is
worth the price.

Mr. Lawlor: I think you said that it costs about
$80 to get an abstract prepared to feed into the
machine. According to my calculations, we could
also punch tape or punch cards of a 13 column
patent for $80 and then it also is ready to put into
a machine. Now actually your costs of $80 are
low. And I think that as soon as you get this proc-
ess speeded up, which it will be, I am sure within
two years, the cost of abstracting will exceed by far
the cost of recording full text.

Mr. Pfeffer: Of course, we also hope that the re-
search in pattern recognition will come ahead to the
point where it will simplify this matter and decrease
the costs, too. But the important point here is that
even if we could get full text into a machine cheaper
we still do not know how to utilize it. Mr. Rock-
man submitted that the examiners were best suited
for analysis work since they had the motivation of
using the system and had the knowledge of the art
and their own experience upon which to rely. In
addition the examiner could perform similar analy-
ses on his pending applications, so that on issue of
the patent no further work would be required. Mr.
Pfeffer agreed that past experience showed that the
examiner who used the system had the greatest
motivation for producing a reliable analysis.

Mrs. Mar den: I think when we consider having
more analysts analyze an increasing body of tech-
nical literature, we are beginning the consideration
of the problem in the middle. It has been esti-
mated that within about 20 years, the time of all of
the world's scientific community will be spent
analyzing this increasing body of technical literature.
In a way the problem resolves itself then, because
there will be nobody to write the reports, notwith-
standing the increase in population.

So it seems to me that there are two avenues of
escape from this dilemma. One of them, of course,
is to have mechanical analysis of all of the technical
contents. Even there we cannot escape the intel-
lectual effort of planning systems to have machines
do this analysis for us. There is no way out of do-
ing the work. But an even beUer way is to deter-
mine at the source, where the material originates,
the kind of format which will enable us to have the
information in machinable form. And that is
where I think we ought to start consideration of this
problem.

161

Mr. Rotkin: I would like to take off from what
Mrs. Marden said. In this sense, there are two
problems which have been mixed here. One is the
problem of today, the other is the problem of to-
morrow — we will have machines tomorrow. I do
not know what kind of machines. I do not know
how they will work. I do not care. But I do know
this. Today the techniques of classification, though
they vary from country to country, are all an in-
dispensable tool in the machines of tomorrow. No
one has indicated that a good classification technique
today will hinder a machine technique tomorrow.
On the contrary, it will help. So it seems to me
that if we enlarge on the classification effort today,
we are helping the machine tomorrow. And what
I would like to say with respect to Mr. Pfeffer's
remarks, is that at least in the American Patent
Office, the technique of shifting from the position of
being an encoder today, a classifier tomorrow, a
part-time examiner the next day — I don't mean on
a daily basis, but for reasonable periods of time —
seems to be working out when it is not drowned in
the flood of expediency of getting rid of a backlog.
In other words, if you try to work at a program
where you examine to become familiar with a par-
ticular art and problems in that art, and then you
study the literature, into the schedule; by the time
you are tired of encoding, you go back to another
project as an examiner. I think this solution would
help today in getting us better classification and
provide a working tool for the machine man
tomorrow.

Dr. de Haan: I quite agree with the former
speaker. I also intended to say that there are two
problems which are completely mixed up. And
this mixing up may result in an error in the auto-
mation of searching. I see the problem thus. You
have one problem which is to put the facts con-
tained in the document into the machine. By facts
contained in the documents I mean the facts re-
ulting from an analysis of documents in a manner
which Mr. Pfeffer pointed out. And the second
problem is how to get the facts out of the documents.
We, in patent offices, all know that the retrieval of
facts from the documents can be done by the
examiner. If we retrieved the facts from the docu-
ments in this way, I am aware that the experts think
that they can bring them into the machine in such
a way as to obtain the answers we need for the
novelty search. But, as to the automatic retrieval
of facts from the documents, up until now I have
learned very few things about systems which might

lead to such results that we could use them to p
the information we want for novelty search into tl
machines.

Dr. Bigelow: The point I wish to make has be<
made partly, and perhaps better than I will make
by the three previous speakers. I would like to s;
that it seems to me that one ought not to be co:
fused by questions as to the availability of low cc
student manpower. This may be a fine asset,
the same sense that having a nice deposit of uraniu
may be a good asset. But it does not affect tl
question of how you design equipment dependii
upon the nuclear properties of uranium. It seer
to me that the question that I would expect we w
be discussing is what is the best mixture of hum?
effort plus machine effort, which will accompli;
the purposes of a patent system. That mixture c<
be zero machines and all human or it could be son
fraction of machines with a differing amount
human error mixed in with it. Now on the oj
hand it is very difficult for me to understand, as
listener, exactly what the context is by which tl
one speaker says we should have everything in tl
document in the machine. On the other hand,
is very difficult for me to understand the context
which another speaker says that we should opera
primarily upon the abstract. I can see situations
which both of these statements are true. I do n
see exactly which problem or class of problems ea(
person is trying to solve in making an absolu
statement of this sort.

Let me go slightly further. It seems to me qui
clear that any operation you do upon a documer
and this point was made, in fact, by the gentlem<
from Western Reserve must be classified as an i
terpretation. You can put the symbols of the doc
ment into punched cards, but you cannot put tl
intent. There is always an operation of intepret
tion changes with time and with style. One a
go back over scientific literature and find things th
the scientific great people of the past said whi<
we did not understand then because they were sa
ing things whose thought context was, say, 20 yea
ahead of their time, and later on we understar
what that person was saying. This is inevitab
true. It seems to me, however, that the centr
problem which we cannot really avoid facing
exactly the problem of considering how to desij
some sort of mechanical aid to some sort of hypoth
sized class of human talent to work with it, in su(
a fashion that the aid in the mechanical system
positive, either measured in terms of per dollar, pi

162

jcument, or whatever metrics you wish. The
ucial point really is to investigate the design or
vestigate the classification system which will lead
the design of some evolutionary process where the
iviously mechanizable features of a patent opera-
>n are carried out, and where, for example, the
d result is that you will either use lower skills in
e final operations or you will use, on the other
tnd, a smaller amount of time of highly skilled
ople. Either of these would lead to an im-
ovement in efficiency. I did not notice people
entioning the possibility of using, for instance,
ry highly skilled examiners operating through
e interpretative properties of the machine and
me classification system associated with the ma-
ine, so as to reduce the number of documents
ey would have to read by elimination. This
:ms to me a very important aspect of the problem
lich cannot be eliminated as part of the process
lich is being studied.

Therefore, what I would submit is that there
her is a case for mechanization or there is not.
le question of how to do the mechanization must/

found upon a careful examination of certain
isses of statistics of present day patent examina-
m use. There must eventually be developed esti-
ites of confidence and reliability in the class of
erations which we are trying to do effectively,
lat is, there must be estimates of how effective and
w precise and how close in agreement a given
Dup of patent examiners now are. Then some-
lere there must be a price attached to this esti-
ite of error. When a price is attached to this esti-
ite of error, then it is possible to conceive of a
ichine and conceive of the expected error it will
oduce on some part of these operations, and the
ice that it costs you to do that. But without more

less this sort of a framework, it does not seem
me that one can say absolutely that all of the docu-
:nt must be in the machine at a certain level. It
es say that for certain classes of situation, you
11 certainly have to see all the documents. There-
re, the question of how to partition this process off
a free variable for the design of the man-machine
item. One does not throw out all the documents,
I does one look at all of them in doing a search
r retrieval purposes in all of their context.
There may be cases where you have narrowed the
Id down to a small subset of documents, and for
ese, 100 percent reading by a highly skilled person
the only thing which will answer the question,
atistics are needed on these questions. How much

time of a patent examiner is spent in extremely care-
ful reading of a small set of documents, compared
with how much time he spends on glancing rapidly
through classes of documents which he can dismiss
by the first, second or third paragraph or a very
cursory reading. I, for one, do not know what these
distributions look like. It seems to me we ought to
be able to get at the data somehow.

Mr. Jeffery: I would like to strike a note of opti-
mism here. In Australia we have been analyzing,
encoding and searching patent specifications in 26
different classes since 1946, and we are still rela-
tively sane, I think. And the reason for our alledged
sanity is that we soon came to realize that one can
overdo an analysis. We came to realize the terriffic
resolving power of a few broad descriptions, as you
call them. Just to give an example, take the whole
civil service of the United States. Let us make a
search using four descriptions first of all red hair.
Red hair is a broad description, isn't it? But that
cuts down the civil service of the United States by
about, say, a fourth. Then we have messenger
boys — redheaded messenger boys. That cuts it
down still further. Ages 16 to 18. In Washington.
Now use four descriptors in my search. Each of
those is quite broad, and yet I have come down to a
fairly reasonable number of items. If you divide
a mass of material up into half and then half and
half and half, you only need 20 of those operations
to reduce that by one millionth. So I think perhaps
we have been mislead in our analysis by thinking
that we have to represent our document all over
again by things we call descriptors. We are trying
to describe the document. I think that word de-
scriptor has mislead us. We are not there to de-
scribe anything. We are there to retrieve. I would
rather have the word retrievor. By that approach
we find that we can just use very, very broad terms
that are immediately obvious reading through the
specification. The analysis is quite rapid. Any
specification that has more than about 20 descrip-
tors or terms associated with it is a big one in our
way. So that is our point of view. We did start
off, of course, by trying to be precise and fine and
get the very thing out again, but we soon learned.
It just broke down on us, and then we came to this
wonderful discovery, of course that 2 20 equals 1
million. But there we rest and we found it quite
successful.

Mr. Fairthorne: A small comment on terminology.
The inventor of the term descriptor. Calvin

163

Mooers, recently remarked it would have been
better called a rejector, which supports Mr. Jeffery's
very profound remarks, that when you are describ-
ing a thing, you are not doing the same as identifying
a thing. When you are identifying or describing
things, you are not necessarily doing the same as
retrieval things. And quite different means are
required for all three functions. Description is
one thing. Identification is another, and retrieval
is another. And the question of descriptors is, in
fact, raised very intelligently in terms which now
could be put in information retrieval terms, in a
footnote of David Hume in his essay on "The
Populousness of Ancient Notions," which refers to
why certain terms are good descriptors or rejectors
and others are not to which I recommend to your
attention.

Mr. Roller: I would like to add another one of
Mr. Jeffery's hopeful notes to the question raised by
Professor Bigelow. It has been recognized that
there is a balance, you might say, between the abili-
ties and the work assignments we would like to give
to both humans and machines. And we would like
to find the best fulcrum point so that we can make
the optimum use of the abilities of both. I think
Mr. Fairthorne will support the position that in
terms of present technologies, the machines are ex-
perts at clerical operations. And almost any job
that we can define in terms of a routine, repetitive
kind of operation, I think we will have to call clerical
in this context. What I would like to mention in
partial answer to the questions you have raised is
that there are some experiments which are under-
way both in the United States Patent Office and
in one commercial organization in Europe, attempt-
ing to find this best balancing point, where the job
of analysis is being divided so that the human being
performs what we can best recognize as being an
intellectual operation, and all those clerical parts
that are normally done by humans are then shifted
over to the machine. Some of these things that are
being conceived for the machine in this way are
going to be quite tricky to work out and yet we can
see that the burden on the human which is going
to have to be with us for a long time, can be
minimized far beyond the point at which it is now
being accomplished. We can talk about that later.

Mr. O'Day: I would like to strike another note of
optimism if I may. This time about analysts. I
have had personal experience with more than 20
analysts over the past year and a half, who have

analyzed in considerable depth in some paten
I find that the good analyst does exist. The pers
who can work 8 hours a day in great depth a
concentrate without too much harmful effect
accuracy and quantity, is an actuality. A nuck
scientist is hard to find. A good examiner is ha
to find. This is true also of good analysts, but th
do exist.

Dr. Bigelow: I would like to ask Mr. O'Day if
would add to his description that he would like
provide them with good tools.

Mr. O'Day: Yes, they must have good tools.

Dr. Bigelow: And I would like to add the followi
as sort of a footnote. It seems to me that when
anybody who tries to imitate a human being w:
a piece of machinery belongs in a psychiatric ins
tution, that nevertheless it is also true that one oue
not to be happy with a situation in which the pi
cedures which an analyst applies are somethi
highly subjective to him. They are, in fact, ;
tributed to him, but to the extent they can be coi
municated. It is necessary in running an organh
tion of running a large complex of men a:
machines, to try and get and to try and force ir
standardized statements, into understandable sta
ments (understandable in the sense that they can
interpreted by the other workers in the field ) all
the tools he uses. Some of the remarks which w(
made, which are interesting and true, I believe, cc
cern the best person to prepare a classification s 1
tern or to prepare an abstracting system being t
man who is going to use it. They are fraught wi
a certain danger. In the first place I do not s
how this man is going to know any more in a str:
sense about what he will need ten years from nc
than his isomorph, some analyst. I question tl
point. But let us waive that. But even if he di
the difficulty is, as you will find yourself, with
specialized subjective art, and not a sciem
Therefore the importance of trying, striving co
stantly to reduce to an attempt at formalization
the various things he is doing when he interprets
document at each stage, must receive special ei
phasis. As this formalization becomes successfi
it automatically follows that he can mechanize pa
of it, so that one will constantly gnaw away at tl
problem of what an analyst does and he will co
stantly move further out to the fringe of uncertain
and a wider spectrum of interpretive acts, t
routine acts being constantly taken away from.

164

lis is the kind of a process which I envision, which
seems to me very crucial to try and prepare for.

ir. O'Day: We have not always found that exam-
lers are the best analysts. Many times they at-
:mpt to read themselves into a patent and say
Well we do not need this. We do not want this."
nd they will violate our very critical rule in the
stem.

(r. Frome: Dr. Bigelow, we have attempted to do
cactly what you suggested. In fact, we have
:ached the point where we use human beings right
dw for what they can do best. Let us take the
:latively simple task of coding and recognizing
lemical compounds. A human being reads a
ocument and recognizes a chemical compound,
rom that point on, the machine can completely
,ke over by means of dictionary look up. That is
hat is in fact being done today. The only thing
e are using the human being for is his associative
Dwers, his ability to read and recognize a chemical
>mpound. From that point on, it is mostly
mtine and the machine can do that. Now some
ay we hope that the machine will be able to recog-
ze a chemical compound in the text. But that
ay seems to be a long time off.

I. Bigelow: Is it likely that the definition of a
lemical compound will shift in time so that this
atement will never become true?

\r. Frome: That is right.

r. de Haan: Mr. Moderator, to verify how diffi-
llt the analysis of documents is sometimes, I might
11 you a personal experience I had in our Board
f Appeals. In the first instance, an application
ad been rejected on an anteriority, which was
most exactly the same as the inventor's. Then the
ventor came in to appeal with a very clever attor-
*y. The attorney proved by analyzing the docu-
ient completely, sentence by sentence that the
athor of this anteriority had meant something
uite different from what the inventor had meant,
ut all the same, if you had read this document
ith normal attention, you would have come to the
inclusion that he had said exactly the same thing.

ir. Stein: I would just like to mention some obser-
ations that we have made in connection with the
nalysts on our electronic research project. We
ave found that there is, statistically speaking, a
reat discrepancy between the different analyses
'hich are performed on the same document. And

we have also found that the same person will take
the same document and analyze it differently on
different days. However, we have observed one
very interesting phenomenon. That is that the
important aspect of each document, the inventive
aspect, the heart of each document, the major
terms are nearly always analyzed correctly. And
it is rather the peripheral terms, the secondary terms,
which are sometimes correctly analyzed and some-
times not.

Mr. Frome: A note of caution on that. What you
call the heart or the essential things of a patent
today may not be so tomorrow. I am talking now
about specific instances. In early references to the
manufacture of polyethylenes, nobody cared
particularly whether titanium was a catalyst or not.
And if someone read it, they would pass it over.
When Mr. Ziegler came along the emphasis sud-
denly shifted very greatly and everybody was
interested whether there was a titanium catalyst in
the patent. So you never know when you analyze
a document what will be important today, what will
be important in the future.

Mr. Kirsch: I would like to pick up a point that
Julian Bigelow just mentioned, that I think needs
further emphasis. I would like specifically to pro-
pose a type of tax on the examiner, as if his load
were not sufficiently difficult already. There exist
today in the computer programming profession some
fairly sophisticated programming languages which
have enormous descriptive detail, and enormous
descriptive ability. It is possible within these lan-
guages, which are quite formal languages and bear
only incidental correspondence to any of the natural
languages that we speak to describe extremely com-
plex processes. I do not pretend to know whether
these processes, complex as they be, are equivalent
in complexity to the type of processes that the
examiner or the analyst uses during the analysis
stage of patent documents. However, it seems
quite plausible that one might undertake to develop
sophisticated, descriptive languages with particular
utility for patent analysis, and then impose upon the
examiner the perhaps minor tax, of devoting a small
fraction of his effort to describing how he did what
he did when he does it. The net result of this would
be something that I think would be rather different
not only in magnitude, but in quality, from what
we have now namely, the ability of the second
examiner performing an examination of a patent,
or the second analyst analyzing a patent to have

165

slightly superior ability to the first analyst. That
is to say, it is certainly within the realm of possibility,
though I know nothing whatsoever about the degree
of practicality of this that one might accumulate
analysis experience within a machine and then im-
plement this accumulated experience in the form of
a machine aid to the analyst, in such a way that
the tremendous efforts that we are proposing to put
into a document analysis will result in some accrued
benefit to us over a period of time, so that we will
get better and better doing something which right
now we do as sort of a one-shot, each job handled
individually, and unrelated to each previous job.

Mr. Van Waasbergen: One of the gentlemen some
minutes ago said that in analyzing certain docu-
ments, it is possible to overlook a certain point,
because at the moment of the analysis, it has no
importance at all. It is quite possible that in the
future such a feature becomes highly important.
But I do not think that this is a very big advantage.
If you make an analysis of a document you do not
make it for eternity, you make it for a certain
period. It is just the same with the classification
systems which are now in use. We all know that
the classification systems have to be recognized con-
tinually because their starting points were not the
starting points we need now.

Mr, Fairthorne: We, of course, always come back
to this question of the stability and consistency in the
interpretation of the text. The fact that people do
agree to some extent (and indeed they do) and that
language does work, seems to be due to the fact that
you always use a document for two exactly the op-
posite reasons. One is that it is familiar and the
other is that it is novel. No one wants to read any-
thing which is totally familiar, and no one wants
to have anything that is totally novel because he
cannot understand it. And those two factors are,
in fact, the cause of stability over a short term within
any particular activity. How to formulate it is an-
other matter. Now, it seems to me (and I had
often been observing the chemical abstracts people)
that this lack of consistency among trained people
is extremely important. It is not secondary. It is

the things on which they agree which are the famili
aspects and therefore would be regarded as co
flicting with novelty in the patent office sen:
Where they do not agree is clearly the novel featui
of the documents they are dealing with, and, thei
fore, those features are ones which want to be ]
garded with considerable attention. Not becau
they agree on them, but because they do not. Ai
this is not hopeless. It does give some idea as
where the future interpretations are likely to
difficult.

Now it is perfectly true as has been said ma:
times, today and by myself at Princeton last wee
and a good many other people, that the obvio
thing to do is to cut out all these difficulties of pred:
tion, which is intrinsically impossible, in spite
the statistician. After all the world is not run
little, green, old men running roulette wheels,
least I do not think so. It looks like it. The thi:
is to get hold of what is written. The only thi:
that is permanent about documents are marks
it. And you want to get hold of that now and int<
pret it for present needs rather than" say now, wh
somebody will want in the future. I know all thi
But at the moment we do not know how to t
somebody to go to the Library of Congress ai
look though every book and recognize what i
want. Until we can do that, we will just have
abstract. That is that. And also, when we can t
somebody to do that, we still have the job of maki:
a machine to look at that text, and in my persor
opinion, that is a bigger job than the intellectual or
I was an engineer once. I was also a compufc
So, this question of lack of consistency, so far frc
being a weakness, is in fact one of the most intere
ing things because a lack of consistency anions
trained instructors is probably giving a very go*
clue as to the elements of novelty within yo
documents.

Dr. Davis: I would not be at all surprised if we a
getting to a point of diminishing returns. This h
been to me, listening as a non-expert, very fruitl
indeed, with significance for the future. Shall j
call it a day, gentlemen?

The fourth session was adjourned at 5 : 00 p.m.

166

Fifth Session

Wednesday Morning, October 25, 1961

The fifth session was called to order at 9:00
A.M. by Mr. Ezra Glaser with Mr. S. N.
Alexander, Chief, Data Processing Systems
Division, National Bureau of Standards,
Moderator.

k/r. Glaser: I am Ezra Glaser, and I am listed as
Moderator. My function this morning is to intro-
luce a different Moderator, who will carry the
aorning sessions. I will moderate the afternoon
essions. We are also adding another short talk,
n the morning, to the program as printed. We are
laving, after Mr. Lawlor's presentation, a short
>resentation by Mary Stevens of the Bureau of
itandards, who will, very briefly, review the con-
ents of a report she recently finished on the state of
he art in machine reading and picture recognition.
rhis is something that came up four or five times
'esterday. People were saying, "Well, some day
nachines might be able to read." Mary will be
ible to tell us, I think, just how much they can
■ead now and what kind of developments are in
he offing. That will be the final item after Dr.
awlor's presentation this morning.

We are also making a modification of the after-
loon session. You will note that there is supposed
o be a roundtable discussion and then concluding
emarks by Dr. de Haan. Well, there are two
hanges in that. In the first place, he will make
emarks but they will not be concluding. They
/ill start the roundtable discussion. I know a little
it about what he intends to say. I think it would
e a big mistake to have him say it and then have
verybody go home because that is the point, I
pink, when you will want to say something. So,
jie roundtable discussion will follow Dr. de Haan.

I will now introduce the Moderator for the morn-
}g session, Mr. Sam Alexander of the National
Sureau of Standards. He is the Chief of the Data
rocessing Systems Division. If I wanted to tell
Jou all of the things he has done in the building of
bmputers and putting them to useful work, it would
se up all the time of the first roundtable. Since
lat looks like a very good roundtable, I will simply

rely on you having heard of him before; his notoriety
spreads wide. I will turn the meeting over to Sam
Alexander, and he will start by conducting the five
man roundtable.

Mr. Alexander: Thank you, gentlemen, it is really
my pleasure. It is a dubious pleasure, on your part,
to have me because I feel that in this area, no
matter how much notoriety I have, it is all out of
proportion to the requirements of the efforts that all
of you are interested in.

I would like to offer a remark to try to put this in
the perspective that I think I see it, and I hope some
of you will see it that way too. In this field of trying
to apply information retrieval to the patent system,
we are faced with the situation that probably would
correspond to an early stage in our knowledge of
childbirth. We have many skilled midwives, and
almost no good embryologists available to help us
with this task, I believe that the general area of
discussion will draw these parallels more sharply for
you. The deep understanding of the process, and
particularly enough understanding so that mech-
anization can be fruitful, requires that scientific
understanding that we so lack today.

I wish I could offer some words of encourage-
ment with respect to the parallel efforts throughout
the scientific community, that might flow in and aid
the effort that so many of you would like to see.
My feeling is that what is being done is very
encouraging, but hardly of the magnitude suitable
for the deep and complicated problem that we have
to unravel. I am sorry that I have to say at this
time, that the situation is that involved and that
complex. I hope that in some of the sessions this
morning you will at least see that there are the
beginnings. I urge all of you, who have a position
by which you can possibly push for the creation of
this more substantial base in my parallel, the crea-

645803 0-63-7

167

tion of the embryology necessary to thoroughly
understand and build up the process into a useful
task, will make use of it.

I hope my remarks will not be interpreted as an
effort to belittle what the subsequent speakers are
about to say. Far from this. I just wanted to
encourage those of you who can help, to recognize
that there is much more, and much more support
and assistance is needed in this field.

With these opening comments, without specific
introduction, since I think most of these gentlemen
have had some participation in the program, I
would like to call on them for their comments re-
garding a look at the patent system in operation.
The first person to discuss this topic is from Aus-
tralia, Mr. Cecil N. Jeffrey.

Mr. Jeffrey: Thank you, Mr. Moderator. Good
morning, ladies and gentlemen. To continue the
metaphor of our Moderator, I would like to say that
"downunder" in Australia we have brought forth
twenty-six young mechanized searches and I think
you may be interested in our experiences.

We have to go back in history — almost to me-
dieval times it seems — to the year 1946 when two
of our examiners, Mr. Hogan and Mr. K. B. Peters-
son, now commissioner, commenced investigating
how mechancial aids could be applied to patent
searching.

The best machinery avaliable at that time was
a Hollerith-keypunch and sorter lent by the Bureau
of Census and Statistics. Starting with no previous
experience, these gentlemen developed the system
wherein one card is allotted to each specification and
is punched at fixed hole positions corresponding
to elementary concepts (descriptors). The first
ten columns of the eighty-column card were re-
served for class and document identification, leav-
ing 840 positions including the X and Y positions.
(See References.) In these references we use the
term "dissective", coined at a time when "descrip-
tor" was not in common use; otherwise I think you
will be able to follow the terminology used.

The first class of subject encoded was electronics.
In three months, the gentlemen mentioned, analyzed
and encoded 2,000 specifications using 450 descrip-
tors ; I think you will agree that this was a magnifi-
cent effort.

Last night I heard a most interesting talk by Mr.
Kalikow at the Patent Lawyers' Club. I was par-
ticularly struck by his statement that he uses 475
descriptors to encode the claims of about 6000

patents on electronic subjects — a figure comparab
with ours.

About 1947 it occurred to another examiner M
P. Purcell, that one could dispense with the sortir
machine and arrange one card per descriptor coi
taining holes representing the documents. Th
you will recognize as the peek-a-boo Batten syster
It came as a surprise to find later that this id(
had been conceived elsewhere independently. B
cause of our investment in the sorting machine:
and the limited capacity of the 80-column car
the Batten system was not introduced on a workir
basis, although it has always been in our though
and particularly recently as I will explain late

The obvious increase in speed of searching cor
pared to our previous hierarchical type of card i:
dex, justified the experiment and whatever tin
could be spared was spent on encoding some of tl
more complex and bulky classes of subject matte
The choice of subjects to be so treated depend(
somewhat on the enthusiasm of the particular inc
vidual for this radical innovation, but generally v
were able to select those classes which were activ
and which promised an early return on the big i
vestment of man-hours needed for analysis ai
encoding.

In the course of our work we met all the usu
problems and made our share of mistakes. V
were fortunate in learning early that descriptc
must be chosen from a consideration of the materi
to be encoded and not pre-selected to satisfy son
neat and logical plan. We encountered false dro
(referred to in our literature as "phantoms"). E
the field of chemistry, these can be troublesome, b
in mechanical and electronic arts they are usual
immediately recognized. We have also met tl
temptation to make every precise searches in tl
hope that the final number of documents selected 1
inspection will be small in number, and by som
times painful experience, are learning that a reliat
selection is one more broadly sought so that it i
eludes a margin of irrelevant material — the "f«
safe" approach.

Our present position is set out in the Table. Yc
will observe there that the number of descriptc
is between 300 and 400 for decks of between 5(
and 6,000 items. There seems to be no direct pr
portion between deck size and the number of d
scriptors for the larger decks. While it is perha
not surprising that, as mentioned above, IV!
Kalikow and ourselves have independently select
about the same number of descriptors for the sui

168

;ct matter it is notable that for different types of
jbject the number of descriptors is comparable.

do not know whether this is coincidental, the
working of some undiscovered law, or simply due
) the fact that the available positions on the 80-
olumn card is of the order of 800 and we are being
onstrained to use only half the available hole posi-
ons by the realization that we must leave room for
jture descriptors.

You may also note from the Table that we have a
lachine search for steroids. This was developed
idependently of that of our friends, here in the
IS.A. The schedule of descriptors is very similar,
ut the points of difference will no doubt give rise to
>me useful discussion between us. This particular
roup of chemicals is particularly suitable for
lechanized searching because it forms a self-con-
uned recognizable group; it is more recent art with
o large accumulation of documents, yet it is very
ctive at present. It does not present too large a
isk and can be undertaken in one intensive burst
f effort. We are seeking other groups of subject
latter with these characteristics and would be
leased to hear of any others you may find.

One is compelled to sound a warning note by
rawing attention to the possibility of bottlenecks
spurring when too many classes of material are be-
lg searched with one machine. I think we are
pproaching that difficult situation where one has
)o much work for one machine and not enough
)r two. The situation is confused by the erratic
icidence of lost time due to machine breakdowns
nd maintenance and by the increasing use to which
le machinery is being put for other purposes of a
atistical nature.

This is one of the reasons why our attention is
^turning to the Batten system which, of course, is
ot centralized at one expensive set of machinery,
'his system too, has the advantage of browsability,
nd enables the examiner to try a succession of
uestions of varying depth and direction in a short
me.

In conclusion, I would like to say that we have
een using this mechanized system for a long time
nd on a variety of subject matter. We have re-
Drded the descriptors used in encoding the specifica-
ons and in searching them. This is a lot of data
Dout descriptors. It should be possible to discover
le frequencies of use of the various descriptors and
erhaps find out why certain ones are more popular
tan others; these are possible principles of selec-
on of descriptors for certain kinds of subject.

Under the pressure of our common needs, and in
the strength of our future cooperation, I feel sure we
will find the funds and labour for this fascinating
piece of research.

References :

Jeffery, C. N.: "Application of Punched Cards
to Patent Searching," Journal of the Institute of
Engineers (Australia) 26, No. 6, 107-110, 1954.

Petersson, K. B. : "Information Systems in Docu-
mentation", Interscience Publishers, New York,
1957, Chap. 22.

Higginson, H. L. and A. Poplawska: "Progress
Report in Chemical Literature Retrieval", Inter-
science Publishers, New York, 1951, pp. 85-104.

Mr. Alexander: Thank you, Mr. Jeffrey. I can
not resist abusing my position as Moderator, to
point out that a large fraction of the American busi-
ness and government public issue pay checks by
punch card and I am sure the phenomena you de-
scribe would be catastrophic, if it were as regular
as you seem to imply. I think one of the troubles is,
that it is difficult to insert the appropriate controls,
which they can do in numerical work.

The next discussant comes from the Federal
Republic of Germany, Mr. Werner Rubach.

Mr. Rubach: Mr. Moderator, ladies and gentle-
men. I will try to give you a short survey of our
documentation activities. The incessantly increas-
ing backlog of pending applications having now
reached 182,000, has compelled the German Patent
Office to consider how the examiners' work could
be made easier. When taking first steps in this
direction, we had to face the fact that the examiners
are very conservative people and do not like a big
change in the search procedure, which differs very
much from the system in use. Having a very heavy
load on their shoulders, and being expected to finally
dispose of at least 1 1 cases a year, it was not found
possible for the examiners to spend considerable
time in analyzing specifications or other relevant
technical literature in such a way, that technical
facts of interest in their particular fields could be
easily found if needed. In the beginning, our ac-
tivities were therefore directed toward search ex-
aminers who showed a special interest in the build-
ing up of documentation in their particular fields.
In our activities we were guided this time by the
thought that the examiner should remain independ-
ent of any centralized organization (that means
documentation center), and that he should be

169

Australian Patent Office

Mechanical Searching Files — Situation At November 1961
A. SERIAL FILES (I.B.M. 80-column card. Sinsle-column sorter.)

Category

Chemistry
(organic).

Chemistry
(others).
Electrical. .

Mechanical

Class or
sub-class

09.10

09.12

09.2

09.3

09.4

09.5

09.60

09.61

09.62

09.8

09.9

87.1

73.1

X05.4

Y05.4

X05.5

Y05.5

05.59

06.1

X04.7

Y04.7

61.2

64.7

77.1

70.1

90.5

Subject matter

Oxygen-containing compounds (excluding heterocyclics

and phenols).

Steroids

Oils, waxes, fats

Hydrocarbons, phenols, peroxides, halo compounds

Polymers, essential oils, resins

Dyes, tannins, diazo compounds, brightening agents

Nitrogen-contairling, heterocyclic (3), organo-metallic,

derivatives of inorganic acids.

Phosphorus — containing

Phenthiazines, Phenoxazines, Phensilenazines

Metallic cyanides, cyanimides and cyanates. Organic

nitriles, cyanates and thiocyanates.

Biological products •

Drugs, pesticides, etc

Coating metal by galvanizing, sheradizing, etc

Facsimile

Television

Pulse techniques

Radio navigation, radar, antennae

Radio systems and circuits (other than X — or Y)

Electron discharge devices

Electric control

Capacitors, resistors

Gyroscopic devices

Hot air engines

Locks

Moulding, casting

Aviation

File

No. of

contents

descrip-

(approx)

tors

1,700

(1)34-

(2)1,200

45!

300

(1)34-

1,000

(1)34

2,400

400

(1)34

3,400

200

130

300

(1)34

800

(1)34

3,400

28!

1,000

16j

300

2,000

46!

2,600

46i

2,300

49;

6,600

5,500

45 l!

2,000

(5)36

970

(5)36

430

140

980

2,000

2,100

B. INVERTED FILES (I.B.M. 80-column card. Batten system.) (These are in experimental or testir
stage.)

1. Transistor devices. About 1 20 term cards developed so far. File will contain about 500 document

2. Sub-class 09.62 (see above). This is a concurrent system. The number of term cards and fi

content is the same as in the serial file.

NOTES:

(1) The one set descriptors used for these six classes.

(2) There may be more than one card per patent specification.

(3) It is proposed to remove heterocyclic compounds to a new sub-class and develop a separate code

descriptors for them. Serial file system will be applied.

(4) This class is also searched by Batten system concurrently.

(5) One set of descriptors.

170

nven a tool which would enable him to have rapid
iccess to any information which might be hidden
anywhere in his search files. This tool should be
}f such a kind, that it could be handled from his
Dlace of work. Furthermore, we had to keep in
nind that an examiner, on the average, has about
15,000 to 17,000 specifications, including other
-elevant literature units. We therefore thought that
n the beginning, this tool could be a set of hand
Dunched cards. In fact, we use a punched card
laving thirteen rows of holes. The holes cover only
lalf the space of the cards, so that we can bring in
;he interesting technical concepts, facts, data or
descriptors, in natural language and additionally
;he most important drawings. Having drawings at
lis disposal seemed to be important, because in the
nechanical engineering field and in chemistry, our
examiner is accustomed to look carefully at the
Irawings.

As to the question of who can do the analysis, we
'eel in our office that the examiner is the right man
:o do this. In order to speed up the work, we have
riven the examiners for his particular purpose,
graduate engineers who do the analysis after having
Deen carefully trained by the examiner and having
3een made familiar with the aspects which are re-
garded as important in the particular field. Before
:he analyst can start work, the examiner is expected
:o set up a list of concepts (descriptors) which in
:otal cover the field on which the examiner is work-
ng. By doing this, we hope to avoid the use of
:echnical terms which might not be quite correct or
which might not give the correct interpretation of
:he information in a given specification. We have
*ven tried to engage students for this purpose, but
ve were not very fortunate with the results. There
ire examiners too, who do not like to collaborate
with the analysts, but who, nevertheless, are inter-
ested in building up a card index in the course of
ime. How can this be done in the most econom-
cal way? We think it is reasonable, for the exam-
ner to record the information he gets from a speci-
ication or other technical literature which he has
:onsidered, when disposing of an application which
le is examining.

In the future, the examiners should not be
allowed to forget the information they have found
when they are acting on particular applications.
When applications for similar inventions reach the
:xaminers' desks after a lapse of time, repetition of
the same search for novelty must be avoided. If
they record the essential information in the literature

which has been considered in conducting a partic-
ular search for novelty and the information is stored,
they will obtain in the course of time, a documenta-
tion for retrieval, I might say, as a kind of by-
product of their normal work which costs only a little
additional time.

As we discovered that the building up of a patent
documentation in a reasonably short time takes more
effort than the German Patent Office can afford
to spend, we have looked for help. This help was
given to us by our industry, which is deeply inter-
ested in the Patent Office being enabled to do its
work properly. An Association for Patent Docu-
mentation was formed, which is presided over by
three men, two from industry and one from the
Patent Office. I have the honor to be the repre-
sentative of the Patent Office. It is a new organi-
zation which is open for everyone and every firm,
which is interested in patent documentation. I am
very happy to say that even our friends from the
Netherlands are represented in our test groups.

The practical work is done in so-called test
groups, each of which is active in a different field of
technique. Until now, fourteen test groups have
been formed. There is one test group which is
concerned with equipment and methods for patent
documentation. It was decided that in every case,
the work had to be started with the setting up of a
list of concepts or descriptors. If agreement on a
list of concepts, (whose number may vary from
under 150 to 2,000) is reached, the members of the
test group begin to analyze the relevant literature.
For this purpose, a format has been worked out by
the Patent Office. The members of the test groups
must put the results of each analysis into this format.
All completed formats are sent to the Patent Office.
The Patent Office is charged with the making of the
master hand punched card. From this card, copies
are made and are given to all members of the
test group.

The German Patent Office is aware that at a
later stage, when enough analysis work has been
done, we will have to think of using machines for
patent documentation purposes. In order to be
able to do this at any time, we have already in-
stalled new equipment, so that with the hand
punched cards, a machine card is delivered which
has all information which is contained in the hand
punched card. For the time being, the machine
cards are only stored, but they are at our disposal,
whenever we intend to make use of them.

171

One further thing I might state. Until now we
did not think that we could abandon the hand
punched cards at all. As I pointed out at the begin-
ning, we would like the examiner to be independent
of any documentation center. In his own search
file, the examiner is expected to make the search
by using his set of hand punched cards. But we
think that there is good reason for making use of
machine cards, in such cases in which the examiner
is compelled to ask other examiners for help. We
have had very bad experiences in these instances,
because the examiners who are asked for help, do
not take much care in answering such questions.
We think that we will get better results, if in such
cases machines are used for providing the answers.
To enable us to use machines for such purposes at
any given time, we have to create a second set of
cards. For this reason, two punched cards are pro-
duced each time, one being given to the examiner,
the other being given to the documentation center.
So you see that our activities in the documentation
field are very restricted ones. We regret that due to
the. heavy load resting on our examiners, we cannot
do more. On the other hand, we believe that the
examiner is the central figure in our documentation
activities. We need him and his help. Therefore,
we should try to restrict his collaboration only to
those things which he can do best. We will be able
to win him to our side, if we avoid doing things
which he may regard as being a sort of secret science.
We should try to give him a tool which is easily han-
dled, but which gives him rapid access to any in-
formation which might be hidden in his search files,
when he wants it. As to the search files, we should
be careful to collect only such material which is
pertinent to a particularly relevant field of tech-
nique. For this purpose, we need a classification
tool which is keen and efficient.

Mr. Alexander; I am pushing the speakers gently,
with the hope that we will have an adequate dis-
cussion period. I think I can already see the seeds
of vigorous discussion in some of the desires at least,
of having both ends of the thing, completeness and
on the desk. This I must see. The next speaker is
from the Netherlands. Mr. P. van Waasbergen,
who represents the International Patent Institute.

Mr. van Waasbergen: Mr. Moderator, ladies and
gentlemen, it is with hesitation that I speak here,
because what I can tell you is not very important.
Nevertheless, I shall try to give you an idea of what
is done in a small Institute, but we do not intend

to remain small. We are convinced that mechan-
ical searching in the future is an absolute necessity, i
There are several reasons for this opinion. In thd
first place, there are economic reasons. In mostj
countries of the western world, there is a shortage of
technical staff. Moreover, the quantity of literature!
with which we have to carry out our searches, grows'
exponentially and this literature becomes more and
more complicated. There are also technical rea-j
sons. It is my opinion that in a short time the
classic classification system will no longer be ade-j
quate, because it does not give sufficient entries intoj
the material. We need more entries and mechan-
ical searching will be a means to help us in the!
economic field and in the technical field. Because!
the fact that the Institute is so small, it is at present,!
absolutely impossible for us to make use of compli-j
cated electronic machinery to do our searching.!
Nevertheless, we are of the opinion that it is neces-|
sary to do something.

One of the reasons is a fact which was just men-
tioned by Mr. Rubach, and that is the attitude of!
many examiners. It is a curious fact, that many,;
examiners who day after day, have to handle the
newest and most modern things, are so conservative
that they do not like to change their methods of work.
And if at any given moment, new machinery andj
methods are introduced on a large scale, it will be
very difficult to obtain acceptance if the field has not
been prepared in advance. We must therefore now
make the examiners acquainted with the idea that
other methods of search will be necessary. More-
over, we are of the opinion that with very small, very
modest means, such as hand-sorted cards, it is pos-
sible to get good results. Then there is the last point.
So long as we do not have mechanical analysis of
documents, it will be necessary that documents be
analyzed by men. Then, it does not matter very
much, whether the results of this analysis is applied
to hand-sorted cards, to machine-sorted cards or tc
electronic machinery.

However, by making use of very simple systems.
we can readily obtain a certain amount of experience
in the analysis of documents. It is for this reason
that some years ago, a decision was made to carry out
some experiments. We discussed this question with
the authorities of the Netherlands Patent Office!
You probably know that we make use of the same
search material. It is of great advantage to us that
the Netherlands Patent Office gave us that possibil-
ity. It was then decided that the people of thel
Netherlands Patent Office would work out a system :

172

dealing with carburetors, making use of punched
cards, and necessitating the use of a rather sophisti-
cated machine, and that the Institute would carry
out a number of experiments with hand-sorted cards,
peek-a-boo cards and IBM type punched cards.

The first system we made dealt with electro-
luminescence. There are not very many patents on
electro-luminescence, but they are extremely dis-
agreeable to search because you have to read them
from the first word until the last. It is quite possible
that the anteriority (anticipation, Ed.) may be a
single word which you find on page three, line thirty-
seven, between brackets. So it was necessary to de-
vise a scheme of features which is rather detailed.

The punched card we made use of is a commer-
cially available card which gives us the possibility
of describing 762 features. Our system needs about
670 features. The time which is necessary for an
analysis of a document was, I believe, between 35
and 40 minutes per patent. That does not say that
if you take an average patent, it takes only 35 to 40
minutes to analyze it, because we found that 200
cards referred to 600 patents. That is to say, that
there is an average of three corresponding patents
to one subject. So the time of analysis has to be
multiplied by three.

The practical experiences we had with the system
were very good. We tested it by repeating the
searches of cases we had already searched manually.
A.11 the anteriorities we had found before came out
and also some others. Then the decision was taken
to go on in this field and the system was enlarged
:o the whole field of luminescence.

Meanwhile, the examiner who did the work had
[earned so much about it, that he was able to set
up a system which was a bit less complicated. The
time of analysis was reduced to between 20 and 25
minutes. I think this is a confirmation of what I
said in the beginning, that you must work in the
field to learn about it. Just as in learning to swim.
You cannot learn swimming by reading books and
talking about it. You must go into the water, how-
ever disagreeable it is at first. This system for lumi-
nescence is not yet entirely ready. But the experi-
ence we had with the electro-luminesence was such,
that two other examiners came up with a proposal to
make an analogous system for the dyeing of textile
naterials. We now have about twelve hundred
:ards on the dyeing of textiles, which refer to about
3,000 patents. The field, as you probably know, con-
tains many publications which are very complicated,
and we found that the time of analysis was between

fifty minutes and two hours. Do not forget that
this time has to be multiplied by the factor which
indicates how many patents refer to the same sub-
ject.

This experience, Mr. Moderator, refers to the
normal hand-punched cards. At the same time we
set up a system with peek-a-boo cards, to get experi-
ence with it. We have made use of a commercially
available card, which gives us the possibility of re-
cording 9,000 documents on it. This system refers
to laminated materials. Laminated materials, of
course is a very broad subject. On one side of it you
will find plywood and you could even say that
the transistor also is a laminated material. We
did not however go that far. Nevertheless, the field
we are covering is very broad. The time of analysis
varied between 25 and 30 minutes.

I must tell you that in all the things we have done,
it was the examiners themselves who did the analysis.
We also had the experience that it is absolutely
impossible to do this work eight hours a day and
five days a week. A man could do that for a week,
but then to maintain his peace of mind he had to
change over to other work. So what we did, was to
give a man normal searching work on applications
and a part of this time was utilized for analysis.

Of course when a man found certain anteriorities
in his search, that is to say he had read these anteri-
orities very carefully, he immediately applied his
knowledge to make an analysis and to help put it
on the cards.

Another experience we had was that we had enor-
mously underestimated the work which must be
done between the analysis of the documents and
the final card. It is still a bit curious for me that
the punching of the cards and controlling of the
cards and the writing of the number of the document
on the cards takes a lot of time and we have found
that a very embarrassing thing.

We have also had some curious experience es-
pecially with this system of laminates. We had
two examiners working on the same thing, so of
course, we made a comparison. The first thing I
found out was that one examiner found many more
features than the other one. About twice as many.
Then we found that they did not always find the
same features. This is a question of appreciation
and in my opinion this is an extremely difficult
point. It is necessary to define the features you are
looking at in a very careful way and even if -you have
done that, then there is always the possiiblity of a
difference of interpretation.

173

Now I want to say this gentleman, that I am
not afraid, that I think it is not so bad if at a certain
time one or two features are missed, because I am of
the opinion that in a normal search there are many
more features that are missed. I think that even
then, the quality of the mechanical search is still
better than that of the hand search.

Another experience we had, for instance in the
field of the laminates, is that it is superfluous to
analyze all documents, because there is a lot of
literature which is practically never consulted.
Most of the searches in this field are based upon
materials which did not exist before World War II,
so we found out that it was more economical not
to analyze the old literature and for the few appli-
cations, and for the few searches which necessitated
the search in the older literature, to apply the old
conventional way.

Another experience we had was that you can also
go too far with the analysis. I do not think that
a system is good, which gives you only the reply
yes or no. Because if you put the question into the
system and the reply is "No, it is new" it is perhaps
a bit paradoxical since you have really learned
nothing. Especially in Europe where our claims are
worded so that you first give what is already known
and then you say what is new, it is not sufficient to
know that there is no complete anteriority. It is
important to know what is known. In my opinion,
therefore, a system of analysis should not go too far.
It should give you those publications which come
nearest to what you are searching.

Another experience that we have had is that we
studied certain subjects in view of mechanization
and when we had studied them seriously, it came
out that the mechanization was not at all necessary
because a simple classification was possible which
was much easier and cheaper.

The last system I shall describe is one which we
are now building up. We make use of the normal
IBM punched cards. For many years, in the Neth-
erlands Patent Office, the searches in the field of
regulation were mechanized in a certain way. The
documents were analyzed and the features intro-
duced into lists, and the searching was done simply
by comparing the indexes on the lists. The disad-
vantage was that only one search at a certain time
could be carried out and moreover that there were
many errors made in consulting this list. We are
now putting these same features on IBM cards,
with the advantage that the search can be carried
out much quicker and without errors. Moreover,

we have another idea. We are of the opinion that i
publications relating to regulation are not only im- {
portant for regulation, they are also important for
measuring. If you have a regulating system, you J
have at first a measuring instrument, and you com-
pare the result of the measurement with the stand-
ard, and you do something with the difference.
That is the principle of regulation. Now we have
found that many anteriorities concerning measure- j
ment can be found in the field of regulation. By j
putting the information we have on the punched
cards, it will be possible for us, in a very cheap way
to eet this information.

Mr. Moderator, these are our experiences and
just as I said, I have not very much to tell ; it is not
very important ; nevertheless, for us it is important, j
We consider our experience as favorable. We are!
certainly going on and we will try to find other
fields where such simple means will be of advantage.
Nevertheless, we considered it on the whole as a
preparation for the big mechanical search tools
which are being developed. I thank you for your
attention.

Mr. Alexander: The next person to discuss this
topic is from the United Kingdom, Mr. L. F. W.
Knight.

Mr. Knight: Mr. Moderator, ladies and gentle-
men: whilst I intend to be very brief, I am in a little
difficulty in knowing where to begin. Perhaps the!
right place to begin, in our case, is fifty-six years ago
when our search started. Because in the United
Kingdom Patent Office, we have always had, sincej
our search started, a system of visual signals, which, 1
I think is unlike any system in any other patent
office. The visual signals are clips which are at-'!
tached to cards, the cards being double foolscap size,
and carrying a copy of the printed specification, the
drawings of the specification pasted on the cards, the
abridgment of the specification and various other
aids to searching, apart from the clips themselves.
The clips normally occupy the top of the cards and
their position and/or their color indicates a feature
described in the specification. In many classes,
the clips represent features of the disclosure looked
at from different viewpoints, or representing differ-
ent types of feature. Consequently, it is frequently,
possible, in some classes always possible, to use coot
cept coordination to restrict the search to those cards
which carry a specific combination of clips identified;
either by color or position or both.

174

Normally, up to 180 different classification units
or descriptors may be represented on a single card;
and higher levels of classification may be repre-
sented by color independent of position, or position
independent of color. Consequently, far from look-
ing at every specification in the given file, the exam-
iner for the last fifty years, has only needed to look
at those cards bearing the appropriate clip com-
bination, thus frequently reducing the number of
:ards he need look at to a very small number, and
in many cases, of course, to zero. If he does have
to look at several cards, he may first glance at the
abridgment, and this frequently results in his not
having to look at the specification itself, even if the
:ard bears the appropriate clip combination.

Thus by visual aids to searching, a British ex-
aminer, I venture to suggest, usually needs to look
at far fewer specifications than his examiner coun-
terpart doing a normal search in other countries,
through a file of equal length. The great dis-
advantage of this system, of course, is the space it
occupies. Probably, for every specification we
utilize five times as much space as the space utilized
in the American office for the same specification.

Nevertheless, even this amount of searching has
been considered irksome, and we have experimented
with other systems employing greater degrees of
mechanization. In particular, we have carried out
several experiments using the Batten or "peak-a-
boo" system. An early experiment was undertaken
in the field of welding and soldering fluxes. In this
case we had a very large number of descriptors and
a reasonably large number of specifications. Un-
fortunately, no sooner had some progress been made
in carding our information than the input of patent
applications in this file dried up completely, and it
has been of little if any use to us. Much more
successful experiments have, however, been carried
out. In particular, in 1952, we carded on the
Batten system, luminescent materials and it is pos-
sibly unfortunate that the International Institute
and ourselves should have gone into the same field.
It might have been far better, I think, had different
offices experimented in different fields. But here
we have a file of 1,500 specifications coded on the
Batten system using forty descriptors. We found
that there was an average saving of forty minutes
per search, by using the Batten system, compared
with our combination clipping system using the
same descriptors, but it was found that the system
could only usefully be employed in about fifty per-
cent of the searches needed to be made. That, of

course, was owing largely to the very restricted num-
ber of descriptors used.

Another successful experiment more recently was
in the field of packaging, where a far larger number
of descriptors was used on a file containing some
6,000 specifications. But again we find that only
about half the searches can be made using the Bat-
ten system. But where it can be used, the search
takes approximately five minutes compared with
two hours using our normal clipping system.

A further experiment has been begun in the field
of cathode ray tubes and other experiments are en-
visaged in the very near future in the field of sound
recorders and reproducers and in the field of woven
materials. Steroids have also been under consider-
ation, but since the United States Office. has already
adopted a system of mechanical searching of ste-
roids, I should like to have a look at that before
we go any further with our suggested Batten system
for steroids.

We think very highly of the Batten system and we
are going to carry out quite a number of further
experiments for we feel that it can be used in a
very large number of fields.

Another system with which we have experi-
mented is the use of statistical punched card sorting
machines. A very early experiment in this system
going back to about 1948, was in the field of alloys,
using some 1,600 descriptors. Unfortunately, this
experiment foundered because with the apparatus
then available, after every seven or so searches it
was necessary to reproduce the cards. They just
would not stand up to the wear. Consequently,
we discontinued that experiment because no other
suitable sorting machine was available at the time.

Recently, however, we have had made available
to us the use of a computer operating with punched
cards and we are now using this for searching in
the field of coating with metals, using some 160
descriptors, where we find that although there is
a definite saving of search time, the saving is not
very great, since we can do very rapid searches in
this art using our combination clipping system.
Also for purposes of comparison, we have extended
the experiment of using the computer to "Packag-
ing" cases, so that we now can search any packag-
ing application either by the computer, by Batten
punched cards, or by our combination clip system.
But it is too early as yet to assess the result of this
experiment.

Finally, we are carrying out an experiment using
Findex punched cards, the knitting needle type of

175

system, in combination with positive microprints
mounted on the Findex cards. A positive micro-
print of every page of the specification, of every
drawing, and of the abridgment, is mounted on a
single Findex card. This experiment is being car-
ried out in the field of selector and other controlling
means of variable speed gearing. There are about
2,500 specifications concerned and we are using 72
descriptors. The experiment has now reached the
stage that about 95 percent of the material has been
photographed and the punching of the cards and
the mounting thereon of the microprints is pro-
ceeding rapidly and we hope to have the complete
set ready early in the next year. But this experi-
ment is suffering at the moment by our not being
able to find a really suitable microcard reader. We
can not, therefore, as yet assess the result of this
scheme.

In addition, I should point out that the classi-
fication of every British patent specification for the
past fifty years, is being recorded on punched cards.
This is about 97 percent complete at the moment
and is due to be completed by the end of the year.
This is being done largely for statistical purposes
and for the production of file lists corresponding
with every classification mark, for sale to the public.
The lists as far as we know, are accurate, and are
kept up to date. When this system was started
some eight years ago, no possibility of using these
cards directly for search purposes was envisaged,
but we are now looking into the possibility of using
them in some way or another: it is however unfor-
tunate that a system whereby the cards could be
sorted more readily and in a greater variety of ways
was not adopted.

Finally, I would like to agree with Herr Rubach
and Mr. van Waasbergen, that one of the principal
difficulties we have to overcome in mechanization
proposals is the inherent conservatism of patent ex-
aminers.

Mr. Alexander: Gentlemen, the coffee pot is bub-
bling vigorously, and rather than have a strong brew
that will irritate tempers and lead to intemperate
discussion, I recommend that we have a break in the
presentation and have the last presentation after the
coffee break and the discussion following.
A short recess was called by the Moderator.

Mr. Alexander: Gentlemen, may we reconvene?
We have the last speaker, Mr. Herbert Roller from
the U.S. Patent Office, who I believe has not had
the opportunity before to participate directly in the

proceedings. I would just like to say that he is an '
active member of the R. and D. group, and in par- i ■
ticular has been working on a chemical search pro- ,
gram with our people at the National Bureau of j
Standards in a team effort. His place on the pro- !
gram this morning is to give a review of the re- '
trieval work at the Patent Office. He has kindly i !
agreed to hold his remarks to a rather brief listing < i
of the subjects. The people responsible for these ac- j i
tivities are in the room and you are invited to go ! \
deeper into the subject during the discussion.

Mr. Roller: We have heard descriptions of various i j
approaches to the problems of mechanization from j |
the several speakers who have preceded me. We'j
have seen a great deal of difference in the approaches ;
that various patent offices have taken to these many i
problems. I think that we must attribute the differ-
ences in approach to the differing philosophies re- j
fleeted in the patent laws of the various countries.
In addition, we have such problems as the relation-
ships between the size of the work load and the size
of the professional staff in each office. In general,
Mr. van Wassbergen expressed a good deal of the
philosophy that has motivated our group in the
past, with respect to the things that give rise to the
need for mechanization, and with respect to thei
intensity of need. By way of example, in our Office,
we have to adopt the attitude that every detail of
every published document, may be of equal value
in searching. This is due to various features of our
laws. Also, there must be recognized the fact that
patent specifications, which form the largest readily
available searching material collections, differ in the
various countries. For example, in certain of the
continental systems, you find that typically a patent
specification is restricted almost entirely to one sub-
ject, whereas, according to the custom that we fol-
low, we permit great latitude in the scope of the gen-
eral discussion in the body of the specification, and
concentrate most of our efforts in seeing that the
claims are tied down very closely in scope to the
actual invention. As a result, a single document
may cover an enormous range of subjects. Because
of the value that is placed on all of the subject mat-
ter of any document, we feel that large scale mech-
anization will require extremely deep analysis and
indexing of the subject matter.

The work in the United States Patent Office was
briefly reviewed by Commissioner Ladd, and this
was supplemented by the remarks of Commissioner
Watson on the first day of the meeting, so I will not
attempt to go into too much of past history. I would

176

like to say, though, that at approximately the same
time that our friends in Australia were starting to
work on these problems, the United States Patent
Office was also beginning to work on problems of
using machines to pull us out from the crush that we
could foresee from the continued use of manual
methods of searching. The earliest work was begun
in 1946 and we have been investigating mechanized
searching problems since that date, more or less con-
tinuously.

The points that I would now like to mention will
fall into two classes. First, a brief mention of three
systems that our examiners use today, and second
a listing of the various research and development
projects that are now under way in our office.

We have heard the steroid searching system men-
tioned several times in this conference. This system
has been in use by the examiners since the early fall
of 1957. This system has been implemented in part
by cooperation with an industrial group, an associa-
tion formed of the various American pharmaceutical
manufacturing companies, who underwrote the
prepartion of a good deal of the non-patent litera-
ture. In recent months, this work has been taken on
by the Patent- Office. In this system a direct coding
method is used, each punch position on a standard
IBM card being allocated to a specific structural
feature of a steroid compound. The large nucleus
that characterizes these compounds is assumed, and
the card records all variations and substitutions on
this nucleus.

A second system which is in actual use is one which
deals with organic phosphorus compounds. This
system exists in two versions, one of which uses an
electronic computer, the other of which makes use
of an IBM card sorting machine. The two are not
exactly coextensive in the subject matter covered, nor
in the logical facilities that they provide, so that in
part the two supplement each other, and both are
in use by the examiners.

A third system which is in use was described by
Mr. Stein on the first morning of the conference.
This deals with transistor circuits. The file now is
approaching about a thousand documents in size and
will be roughly double this size when it is completed.
The descriptive factors are recorded for each of the
circuits that are in this file, including some of the
topological features of the circuitry and certain of
the general behavior characteristics.

The rest of our program is still in the experimental
stages and deals with a variety of subjects. In one
project in the mechanical arts, a system is under

development for recording and searching hydraulic
circuits. This, of course, has some features reminis-
cent of chemical structures, other features reminis-
cent of electronic circuitry and many features that
may be peculiar to the mechanical arts. Symbology
for the mechanical elements in the fluid flow paths
involved are being worked out in such a way as to
permit tracing any of the operational sequences that
occur in the hydraulic circuits.

Another system which is under development is one
for recording information about artificial breathing
devices.

Most of the other systems now being developed
deal with chemical subject matter. We were
trapped into believing, as were many people out-
side of this field, that the area of chemistry lends
itself more readily to mechanization than do other
fields. This illusion is largely based on the fact
that there exists what seems to be a systematic tool
to begin with, namely, chemical structural formulas.
However, I think the experience of many people
who have worked with this problem is such that we
realize that this is not as simple as it sounds. More-
over, the experience of other patent offices has indi-
cated that work in fields other than chemistry may
be very worthwhile indeed.

A system which is known as PACIR deals with
pesticides. This also is going to be available in
versions for both computers and punched card
machines.

The HAYSTAQ project is one which the Na-
tional Bureau of Standards and the Patent Office
have undertaken jointly. This system uses an elec-
tronic computer, and deals with chemical subject
matter in terms of compounds, mixtures of mate-
rials, and processes. The largest emphasis in the
last few years has been on chemical structure search-
ing. The system is designed to permit any com-
pound of known structure to be stored and searched
and to permit an enormous-amount of flexibility on
the part of the user in asking his questions.

As a corollary to this, during this past summer
we had a group of students (who were about to
enter the university ) who worked out an algorithm
for taking the material which is the direct product
of a chemist's analysis of a document and manually
recorded in a minimal way and converting this
minimum amount of material into the codes which
are required for storage in this system.

Mr. Stein also mentioned the project which his
group is about to embark on, in which the full text
of a hundred transistor patents will be made avail-

177

able to a computer on magnetic tapes. Various
kinds of linguistic processing will be attempted on
this body of material.

The DRAM project involved an investigation
into the development of an open ended, inverted
file. Descriptors were generated concurrently with
the analysis of the file which was to be mechanized.
Chemical compounds, both organic and inorganic,
were fragmented into substructures which consti-
tuted generic terminology for the compounds.
Fragmentation was done by rules rather than by
the use of a preconstructed vocabulary of fragments.

Non-compound information was analyzed by key
words. A thesaurus-like grouping of these key
words was contemplated. Relationships which
were provided included the roles of compounds and
the association of compounds into process sets.

The analysis was performed sectionally. The
subject matter of the patent was partitioned into
( 1 ) general disclosure, ( 2 ) specific examples and
( 3 ) claims in order to discover various patterns of
information, with the expectation that it would
facilitate the analysis and encoding of the subject
content.

There is other work which has gone on which is
not directly related to searching systems as such, but
which will hopefully, affect the design of large
systems.

A method was developed for symbolizing chem-
ical reactions and processes, which emphasizes the
sites of reactions, that is the atoms and groups of
atoms which are specifically changed, or among
which the connections are changed or altered in
various ways during the course of a reaction.
These are recorded in a symbology which may lend
itself to machine processing. We have published a
report on this work that is available on request.

In another project, a new enumeration and linear
notation system for chemical structures has been
developed, in which a serious attempt has been
made to overcome deficiencies in existing notation
systems. We think that this goal has been largely
achieved. According to the use of this system, every
chemical structure yields a unique cipher, and each
cipher can be translated into a unique structural
diagram. The regularity of this is illustrated by the
fact that in another summer program, an algorithm
was developed for converting the linear cipher for
any given compound into a structural diagram and
displaying it on the face of a cathode ray tube.
This algorithm has not yet been programmed for a

computer, but we have reason to believe that this
will be successful.

Some work has been done on the problem of
storing and retrieving information relevant to legal
decisions. We will hear more on this general topic
later on in this program, but not on our specific
work in this area.

A good deal of work on linguistic problems, which
are at the heart of all of the problems that we are
talking about here, has been supported by the U.S.
Patent Office. Some of it has been performed in the
office. Support has also been given to various studies
in formal linguistics and processing of pictorial in-
formation undertaken by the National Bureau of
Standards.

One additional use of machines that has been
in operation for a number of years at our Office
was mentioned by one of the earlier speakers. For
many years we have had listed on punched cards
(which are available for sale to the public) clas-
sification lists which enable a person wanting to
know what patents are classified in various arts, to
obtain the list of all numbers of patents included.
Fairly recently we have added to these lists the
classification of foreign patents, in the sense that they
are classified by our examining staff into our own
classification system.

We feel that the achievement of a fully mech-
anized information retrieval system for our Patent
Office's use cannot be realized without solution of
many of the basic theoretical problems that exist,
as well as system and hardware advances. It is
anticipated that we will expand our efforts consider-
ably in all of these directions, while at the same
time continuing to make use, in such areas as we
can, of existing techniques.

Mr. Alexander: Thank you, Mr. Koller. Now, I
would like to open the various presentations for
discussion. I hope that in the course of these we can
bring out some comparisons and differences. As
a means of opening it, I would like to offer one my-
self, and that is to comment on the remarks of the
gentleman from U.K. who said that it was unfortu-
nate that they had done the electro-luminescent art,
without realizing that the Institute in Holland had
done the same item.

I think this is a happy circumstance and should
be exploited. The intercomparison of the two pro-
cedures, since they apparently were done quite in-
dependently and I am sure from a different point
of view, would be very desirable. An intercom-

178

parison report I think would be very useful to the
community. I am afraid that in this whole area
there is a great tendency to try to move from an idea
to a useful operation, in a very short period of time
and without thorough intercomparison. With the
wide interest that must exist among all of us in avoid-
ing the same pitfalls, I think cooperative efforts and
intercomparisons would be very, very helpful. I
hope that the discussion this morning can direct itself
toward possible ways of cooperation and possible
intercomparisons of work thus far.

With this opening plea, I am prepared to recognize
questioners.

Mr. Rockman: I would like to ask a question of Mr.
Jeffrey. Do you have what is called an open ended
system, whereby you may add a descriptor to the
system if perchance an inventor happens to come
in with a concept that you do not have?

Mr. Jeffery: Yes, we do that. When electro-
luminescence first came in we added a descriptor.

Mr. Alexander: May I inquire, suppose the item
which has been called to your attention for the first
time because of a claim for novelty, is implied at
least in part, in older documents, would you be able
to go back and up-date the other references? Does
the examiner provide a feed-back as a result of hav-
ing observed that there is a partial relationship in
some of the earlier patents?

Mr. Jeffery: Yes, that is a problem. We have to
be very careful when selecting a new descriptor,
that we make a very broad search in the accumu-
lated material to make sure there is nothing like it.
We make quite an exhaustive search in different
ways.

Mr. Alexander: Suppose there is something a little
like it but not sufficient to completely invalidate all
the newness. So that there is some novelty, and
therefore you must now build back into the old
descriptors, into the old documents, some way to
identify items that were called for but not by this
type of terminology. Do you have a way of up-
dating old material as a result of the fact that the
examiner finds something in the old material, to
which his attention was not called initially?

Mr. Jeffery: No, I do not think so. We would
not treat the old material. A citation is read in the
light of the knowledge at the date of the citation,
not in the light of the knowledge at the time you
happen to find the citation or need it. If the cita-

tion just mentions something in passing, and its sig-
nificance is not realized, well, it is not a citation.

Mr. Rot kin: On that very point, sir, I think this
is at the root of the problem that was discussed in
part yesterday. A man who is looking at a docu-
ment for encoding it, whether he wants to encode
it in its entirety or in part, Is going to encode only
what he recognizes at that time. The fact that he
does not recognize something does not mean that
it is not necessarily in the document. Now, subse-
quent to his encoding that document, something
else may call his attention to the existence of that
facet of knowledge. It may be in the document if
he goes back and looks at it. Should there not be
some technique to permit the extracted information
of the document to be up-dated?

Mr. Frome: Many speakers have said the examiner
wants the output of a system right at his desk. I
do not really believe he does. Our experience has
shown that after he realizes that mechanization is
not a monster, the examiner does not care how you
give him the references. The only thing he would
like to have is fairly fast service, so that after he
has made a search, and wants another one, he can
get it fairly promptly. And it has been our experi-
ence with about eight to ten examiners, that they do
not care whether you use the computer or punched
cards. All they want are the references handed
to them. This illusion that it must be at his desk,
I think, comes down to a psychological fact that
he must have something to do with it. He should
be able to control it. However, if you take the
examiner who is actually going to use it, and make
him part of the system or give him the illusion of
being part of the system, he will completely accept
it. It does not then matter whether he has it at
his desk or not. In certain instances you cannot
put a computer at a man's desk, it is just too big.

Mr. Alexander: Unfortunately, you can put a con-
sole at his desk, but you cannot afford it. So I
think you cannot be quite that categorical about it.
Whether or not this is desirable, has a great deal
to do with the effective work habits of people. And
I think, while I admire your enthusiasm, I believe
there are many facets of this problem.

Mr. Rockman: I would like to get back to the pre-
vious discussion about adding new descriptors. I
assume that the descriptors are listed as generic head-
ings, with others under them as sub-generic, and
sub-sub-generic, and so forth. This would be similar

179

to your classification schedules where you have in-
dented headings. When you have to add a descrip-
tor that you have not used before, if you can fit it
under one of these generic or sub-generic headings
you will not have to look back at every reference you
have under the file but only those that fit under a
previous generic sub-heading.

Mr. Stitelman: We have had this problem and we
have tried to deal with it and we do just that. We
are now working with a miscellaneous group of
transistor circuit patents. Under each generic group
we include a miscellaneous sub-class to avoid search-
ing the whole file when it is necessary to up-date the
file. This in itself will not avoid an exhaustive
search for a new concept, but it should limit the
number of cases where this should be necessary.

Another means for avoiding exhaustive search
through the entire file, for a new concept, as Mr.
Kalikow can tell you, is what Mr. Stern calls the
"peek-a-boo" principal. You would take general
terms, such as time, perhaps or other general terms
which will define a specific new concept. Radar,
for example, could be defined as detection by elec-
tro-magnetic means and echo means. If you use
those three terms, you come up with radar in a broad
sense. Now we tried to include this in our system
as well as specific terms. The examiners in search-
ing will often find that the devices they are dealing
with are defined in a standard way and then there
will be variations in structure later. We want to
give them the facilities to search for such unexpected
variations. This multiplies the number of terms
required and makes it rather difficult to handle the
situation. We have a hundred to a hundred and
twenty-five descriptors for each document, on the
average. This is too many. We will cut that down.
But in any event, a substantial number of terms per
document are required for patent searching.

Mr. Kalikow: I would like to suggest the way in
which we handle this problem in our searching sys-
tem. That is we interpreted or translated every
complex concept, in each of the patents that we
searched or that we worked on in the file, into a
basic language code. In other words, we had no
descriptors or rejectors, that would be complex
words such as spectrophotometer or radar. That
was not part of our rejector code or our descriptor
code. We had only basic language terms. In that
way, we were able not to have to worry about this
problem of up-dating because we always dealt with
basic concepts as we set up our search question and

any new art always took advantage of these basic
concepts. Electro-luminescence, for example, would
never become a descriptor. This would be de-
fined in very basic terms, in our classification of each
of the patents, and therefore, the component mean-
ings of the term electro-luminescence would already
have been classified in each of the patents.

Mr. Jeffery: How would you express a transitor P-N
junction in these basic terms?

Mr. Kalikow: I would have to give you our glossary
where we translated every term, every word of the
English language, into our basic terms. I went to
Webster's Intercollegiate Dictionary and I did this.
You would have to see how this is done. It is a
process that you can train people to do.

Mr. Fairthorne: The belief and the hope, that we
can reduce our statements into basic terms, atomic
terms, a concept of one kind or another is, of course,
a very long one. Leibniz and others tried to use it
in a universal character. It is an unfortunate fact,
or fortunate fact perhaps for the progress of the hu-
man race, that any term is complex if your inter-
ests change. The basic terms even in different
activities are different. The atom of the sociologist
might be a baby and from the physicist's point of
view a baby is certainly not an atomic term. I
think I myself held at one time the view that you
might find certain common interests, shall we say,
in the human activities to be dealt by weighting
these terms or giving orders of preference. I am
afraid I have had to change that view and I think
nearly everybody else has in time.

The reason that anything belongs to a gen-
eral class is not because it in fact belongs to it when
no one is around. The classes and things do not ex-
ist if no one is around. The reason any specific thing
belongs to a general class is because people agree to
say it does, and for no other reason whatsoever. Wc
are rather misled here by the taxonomists of the phys-
ical world and biological world who take objects
and say you are only to look at certain bits and
pieces of them. So in a way you get the idea that
the genera and the species, and so forth, are natural
objects. But they are not. They are interpreta-
tions of them. So, any system which deals with
generic terms and specific terms, just has to have
the lists of them there. There is no way out, because
we do not even have the objects to deal with. We
only have their names. And it is precious difficult to
find out from the name of an object if all the prop-

180

erties in which people may be interested are in it.
You can always make an artificial name if it has
some of them. And when you do that you say you
are dealing with basic interests. So you are. They
are not eternal, and they are not universal. They
are local and temporary, and that is why all these
systems on basic terms work beautifully in small
libraries with peoples' special interests. And a very
good thing too. And I certainly think that excellent.
However, the main problem is how are we going to
make it wider and more stable, and still retain some
degree of, I will not say precision, but refinement.
However, further refinement of course, in the sense
of bibliography and not of social usage. I think we
must be awfully careful when we talk of basic terms
quite legitimately, not to think we are talking about
sort of eternal atoms of the type Leibniz tried and
failed, and a lot of other people have. I know I am
telling nothing new to anybody here, but I just hope
I am putting it a little more clearly perhaps than has
been put recently here. I am certainly not attacking
anybody. But there is grave danger here of think-
ing we have actual atoms of concepts.

Mr. Kalikow: I completely agree with what you
said. It is just that this is a more rudimentary way
of expressing the things. No question but what you
say is absoluetely correct.

Mr. Alexander: May I point out, however that you
find yourself backed into odd kinds of difficulties.
I am quoting a remark by Don Swanson, who when
faced with including the concept of cannibalism into
an already existing scheme of coding, found that it
might be forced in by combining capital punishment
with unusual diets. I doubt, however, if anyone
would arrive at cannibalism via this route.

Dr. Bigelow: It seems to me that one more point
should be appended to Mr. Fairthorne's comment.
Namely, that the crucial thing to look at, in view of
this comment with which I completely agree, is the
question of getting good parameter sets to describe
the effective conventions of usage which are present
today. In some sense or other the whole problem
of being able to automatize some part of the patent
procedure or library procedure, which is now done
largely by humans, depends upon being able to
choose descriptor sets and parameter sets which are
effective for some reasonable time base, such as the
time base it takes you to develop the system you are
working in and to persuade people to use it. And
secondly, that the effective systems shall have some
natural evolution which can be up-dated concur-

rently with the field. So that you see there is noth-
ing like rigor as Mr. Fairthorne said, nothing like
invariance over time and over fields. Nevertheless,
this is the essential heart of the problem. It is also
the essential heart of many scientific problems.

I was thinking, as the speaker before last men-
tioned his classification system, of what difficulties
you get into even within the narrow selection which
might be electrical technology. You have to de-
cide, for instance, whether radiation is a legitimate
word. Now, radiation leads you to certain classes
of equations. One of the big hiatuses in physics at
the present moment, is that, in the small, the equa-
tions are no good. They have been good for at
least a hundred years since Maxwell described
them, but when you get down to small dimensions,
all of a sudden radiation has no meaning. You
have quantities which are related to the old ones
and this descriptor has vanished.

So that I would emphasize again that we have
only relative stability and relative applicability.
What is needed is an effective treatment of these
relatives.

Mr. Alexander: May I ask, does not this imply
then that any attempt at mechanization must have
inherent in it a continuous updating procedure that
does not ground you with clerical effort? I think
the important thing is that the shift in concepts are
gradual. But the tremendous work load associated
with making the necessary clerical notations to re-
cord the shifts, are the ones that usually prevent us
from accepting them as they appear in use. It is in
this area that I think the machines can be par-
ticularly helpful, except that we are not quite clever
enough yet to know how to produce a steadily,
easily up-dateable scheme.

Mr. Leibowitz: I think we are getting down to the
heart of the problem by the recent comments. Mr.
Koller said that the approaches of the foreign
patent offices were different. But I feel that they
are the same in the sense that it indicates that the
state of the art is such that the approach is sub-
jective and empirical. It depends upon choice and
selectivity of descriptors. It is a hunting for what
is the most selective descriptor for a particular
specialized situation. Now, nobody has really said
what these descriptors are. I find that there is con-
fusion in the art, in the lack of definition of termi-
nology. I am very happy to see that they have
given a name to the broad field of information re-
trieval. It makes it easy now to tell people what I

181

am doing. Most of the time that answer, informa-
tion retrieval, ends the conversation.

We talk about analysis. What are we analyzing
for? Has a system been set up by an information
retrieval expert, and then somebody analyzes ac-
cording to this system, or is an analysis of informa-
tion for the purpose of setting up such a system?

There has been a statement made that it is curi-
ous that the examiners are not willing to accept
something new. I do not think that is so. I think
it is a question of their being familiar with what it
is. If they were told what it is they might be quite
ready to accept it. I think the problem is one of
getting all the examiners and all the information
retrieval experts in on the job. Now, if the descrip-
tors are very specific, there will be a wide difference
and a wide disparity in the various systems. As
soon as you go to a higher level of abstraction, you
will find that they are very similar to each other.
However, the job is bigger, because you do not have
much discrimination in very general descriptors.

So, my viewpoint now is, that the best approach
is to extract descriptors which indicate materials
and objects and key words, from documents which
are to be incorporated into a mechanized file as
specifically as possible. Then one can work with
this as a vocabulary and shuffle them around to
generate classes or generic terms. This is modifi-
able as the case may be. With changes in the art,
one can move these descriptors around. One need
not go back to the original document provided one
has been thorough in the first place in extracting
these descriptors.

Mr. Alexander: I think in view of what you have
said, Mr. Kalikow is entitled to a rebuttal.

Mr. Leibowitz: I did not disagree with him.

Mr. Kalikow: No, I do not think a rebuttal is nec-
essary. I would like to point out that my previous
comment about a basic language code, was merely
dramatic for the benefit of someone who asked the
question as to how it might be done in a small file.
Those of you that heard me last night, know that
this is not what I advocate. It is completely im-
possible in my mind, to use any kind of basic lan-
guage or semantic code of this type, to work on the
level of many, many documents as we are working.
This is why I have firmly advocated that we put
the full text of the patent into storage, and that
thereafter we use all of these techniques that you
people are developing here, in retrieval. You can
retrieve on the basis of complex descriptors if that

is more beneficial to you. You can retrieve on the
basis of basic, or more rudimentary concepts if that
turns out to be more beneficial to you in the par-
ticular search that you are making. And you can
have programmed in your computer the relation-
ships among the more complex words and the fac-
tored words. It is very important because these are
the kinds of relationships that must in fact be put
into the computer. But let us not just take one ave-
nue of attack when we come to retrieval. All of
these various techniques that are developing will be
very beneficial, because what we are really trying to
do, is trying to make an analogy in the computer, to
what goes on in the mind of a searcher as he
retrieves from the full text today. We want our
computer to be able to do this and the mind of man
uses many routes for getting the material out of the
file.

Mr. Newman: One warning to people who set up a
series of descriptors in a generic hierarchy. It is a
trick that looks very simple. But one is inclined
to forget that the hierarchy for example of dog, mam-
mal, vertebrate, etc., ignores the fact that my dog
Fido not only belongs to the genus mammal, but
he belongs to another genus called burglar protec-
tion, along with a policeman on the beat and a lock
on the door. As an additional example I would like
to call attention to the fact that a thermometer is
also a heat responsive device and the mere fact that
you do some measuring on it, that you have a scale on
it is incidental. If we go back to what was dis-
cussed on Monday, to the prime thing rather than
the accidental use, it is a heat responsive device and
not a measuring instrument. Using this approach
you will find that you will be able to do better
retrieval.

Mr. Alexander: The nature of this discussion, seems
to bear on the comments of one of our discussants
from the Netherlands. In Mr. van Waasbergen's
remarks, he indicated that he obtained quite differ-
ent results on the same document from several ex-
aminations on this document. I would like to ask
for his comments regarding his appraisal as to
whether this was new since there seems to be this in-
terplay between whether or not the essential content
was left out, or whether he wanted to use a differ-
ent way of describing what he saw.

Mr. van Wassbergen: I can say that both examiners
had the same essential things. The difference is,
and it was rather considerable, that they were, so to

182

speak not in the center but on the periphery. They
were certainly far apart in their descriptions due
to the fact that one of the men was a chemist and
the other one was a civil engineer. The field of
laminates is a very big field, and such things as
plywood and other things in the building field, were
treated by the civil engineer. And the civil en-
gineer, of course, has a different view on technical
things than the chemist. The chemist is accustomed
to work with polythene and that sort of thing. For
instance, the civil engineer, looks to the structure
which is being made with a certain material. So
everything which goes into the erection of the struc-
ture, is called to his attention first. It is difficult
to ascribe it to one thing. There is a whole complex
of circumstances which have to be taken into ac-
count. The best way is to have these people dis-
cuss all cases about which they are not entirely sure,
and then they come to a common result if they are
intelligent people. If they are not intelligent you
must not put them on the job.

Dr. Wooster: I am just wondering if the word de-
scriptor is being used too loosely here. When I use
Descriptor, I use it with a very capital D. I refer
to it as the system developed by the same gentleman,
I believe, who also developed the current "informa-
tion retrieval". His name is Calvin Mooers. I use
Descriptors in the sense of a fairly small set of
terms, perhaps one to two to three hundred, very
carefully defined. Many people apparently are
using descriptors with a lower case d, and it would
refer to anything all the way over to "uniterms."
Some of the confusion that is being incurred, is the
fact that everyone uses descriptors in the same way,
or in a personal way.

One other point which I think applies to classi-
fication systems. A couple of weeks ago I was at
the meeting of the UNESCO Advisory Committee
on Documentation, and we were discussing classi-
fication systems. Madame Bagrova, who is very
bright and the head of the reference department
at the Lenin State library, says that classification
systems are ideological and that they cut across
linguistic communities. We are hoping to get
UNESCO, not in science, but in some of the sub-
jects dealing with "E" and "C" in the UNESCO
title, to try in three different languages to see what
sort of classification systems people will work up.

So we have the personal problem in classification
and each of us tends to classify things differently.
But there is a language community in which we
are working. I think point A is that if we use the
descriptor, we should say in what sense we are
using it.

Mrs. Mar den: Yesterday, Dr. Bigelow made some
very penetrating remarks, concerning the integrated
man-machine use of a system in searching and the
interplay between the two. Until we achieve a
totally mechanized system, it might be well to keep
this idea in our minds, of having the close relation-
ship between the man and the machine. It seems to
me that one of the best ways we can achieve the
maximum utility from this combination is by not
looking at all the material we have in a mechanized
way, but looking at highly compressed stored in-
formation about the material, and then having man
do some searching, perhaps backed up by machine
searching, with the continual evaluation by the man
of the end results. In connection with the searching
of the subject material, perhaps we are trying to
tackle too much to achieve a uniform mechanized
searching of all the material. If we can have spe-
cialized searching according to subject matter, and
have specialized systems with regard to the particu-
lar subject matter we are searching, until we can
have a little more experience with searching the
diverse kinds of subject material, we might eventu-
ally be able to bring all of these diverse systems
together into a uniform system. Until that time,
tackling the whole collection of scientific and tech-
nical material, may be too much for us in the first
try.

Mr. Alexander: My reaction to this is that there is
a story about the Tower of Babel. I am afraid if we
go in diverse directions now, we will never come
back together. This is my fear. Maybe the prac-
ticalities of the situation will dictate this.

Have I overlooked anyone who has not had an
opportunity to have his comments heard? In that
event, in order to give Mr. Lawlor a respectable
opportunity to deliver his remarks, I will suggest
that the corridor discussions take care of the remain-
ing issues. With this I close the discussion session
and I would like to turn the rostrum over to Mr.
Glaser.

64'i803 0-63-8

183

Sixth Session

Wednesday Morning, October 25, 1961

The sixth session was called to order with
Mr. Ezra Glaser, Section Chief, Data Proc-
essing Systems Division, National Bureau
of Standards, Moderator.

Mr. Glaser: Thank you very much, Sam. Mr.
Reed Lawlor is well known to many of you. I am
not going to take any time to give you a detailed
account of his achievements, except to tell you that
he is Chairman, of what I hope to be a very pro-
ductive committee of the American Bar Association,
to advise the Association on the various ways in
which electronic computers and the technology that
goes with them, can assist the lawyer in performing
his functions. You will get a better idea of what is
going on from hearing him, than from hearing me,
so I will turn the floor over to him right away.
Mr. Lawlor: I feel a little bit out of place here.
Except to the extent that I file patent applications
I do not work for a patent office. I do keep some of
the examiners busy and they keep me busy. But
except to that extent, I actually have very little
knowledge of what goes on in patent offices. Some-
times I cannot understand it.

Now, I probably look a little bit like the 1860
medicine man who traveled around with a lot of his
wares to display them, and I do have some things
here which I would like to show you. I will try
to speak rapidly and I will try to cover a lot of
material in the twenty minutes that is available.

I want to open by bringing you greetings from
John Satterfield, the President of the American Bar
Association, and from Joseph Gray Jackson, the
Chairman of the Patent, Trademark and Copyright
Section. I am delighted to be here, and I am thrilled
to hear all of these comments and to hear about the
progress that is being made by all of you. The work
in the American Bar Association is parallel in many
respects to the work that you are doing. It is so
parallel that it is about as incoherent as some of the
work that you are doing. Many of you are work-
ing in diverse directions, but you are coming to a
common end, I am sure. That is also happening
in the American Bar Association. We have a com-

mittee; the committtee is made up of many indi-
viduals, each one of them wears at least three hats
and we all have different interests and different
goals. I won't tell you much about the history
of the committee, except to say that it is traceable
to an ancestor that was born in 1952. The main
funtcion of our committee, as I see it, is to create both
supply and demand. We want the lawyers to under-
stand what computers can do for them. We want
the manufacturers to know what the needs of the
lawyers are.

I would like to talk about two principle subjects
today, though there are many others which are of
interest. I would like to talk a little about retrieval
and I would like to talk a little bit about patent
claims. There is a story that perhaps is suitable for
this audience, and that is the story of the survey
made by one of the computer manufacturers of the
interests of the various people in a corporation. Men
at different levels in the company were asked what
the computers could do for them, what duties could
be delegated to the computer. Everybody agreed,
the President, the General Manager, the Super-
visors. They all came to the same conclusion. "All
the duties of my suborinates can be performed by a
computer, but not mine." And this brings us to the
first slide. ( 1 ) This is a computer here with two
men standing before it. One of them believes in
the abstracting system, and the other believes in the
full text system. One of them says to the other,
"Let me do the thinking around here." That is the
problem that you have. Everybody is in love with
the ideas to which he has grown accustomed. Now,
actually there is a big debate. Should we abstract, or
should we record full text? We have been abstract-
ing in the law for years. Since I do not want to re-
flect on any particular publisher, I won't use names.
But here is a publication in the law field. It gives
the case, the arguments and the facts. At the be-

184

ginning of the case are head notes. These abstracts
are inadequate for the lawyer. But they are his
principal tool. For any mechanical searching sys-
tem something needs to be done either to improve
abstracting, or abstractors. Otherwise we must
record full text. How we will do it we do not know.
The University of Pittsburgh, has been leading the
field on recording full text. They have recorded
all of the Pennsylvania statutes verbatim. There
are six million English words in these statutes. Now
the question arises, what good is it? What can you
use of for? If you had permitted a general lawyer
to index these statutes, he would not have recorded
many of the concepts which are in the text. He
would have ignored them. Certainly if John Horty,
who is Director of the Health Law Center, at the
University of Pittsburgh, where this work is done,
had come across the word patent, or the term letters
patent, he would probably have ignored it because
he is interested in health.

I would like to play a little game with you. I
would like you to write down in front of you, the
number of times that you think the word patent
occurs in the Pennsylvania statutes. And the
lumber of times that you think the phrase, letters
Datent, occurs in the Pennsylvania statutes. And
n a few minutes I will show you the answer. Now
let's take the next slide (Slide A). Lets talk a
ittle bit about what you can do, or what you have
:o do. We hear about this exponential growth of
Information. This plots the number of periodicals,
new periodicals being published each year, as a
function on years. If you plot this on semi-log
scale, you will make an interesting discovery. It is
true. The growth of information in these three
fields is exponential. You get a straight line func-
:ion. In fact, you will find that the three lines are
parallel. The information in all of these fields is
growing at the same rate. It is increasing about
:enfold in fifty years. I do not have similar data
For patents. Undoubtedly you do.

Let's take a look at the next slide (Slide B) to see
what the Congress is doing to us. These are the
lumber of words being added to our Federal
statutes each year. Notice what happened in 1932
ind notice what happened at the end of World
War II. If you plot this on semi-log paper, you
-vill find to your surprise that the rate of growth is
nore rapid than exponential. We need some brave
Congressmen who will repeal laws and replace
them with shorter laws.

<
o

S§

M «

W W

< a,
o o

INCREASE IN NUMBER OF PERIODICALS
IN BIOLOGY, MEDICINE, AND LAW

SLIDE A

Let's have the next slide (Slide C). Now at the
University of Pittsburgh, they use a computer that
indexes the text ( 2 ) . All the words are automat-
ically listed in alphabetical order and opposite each
word such as "incompetent," there is placed a num-
ber which is either the document, or the page, or the
section number, or whatever it may be. This par-
ticular glossary was developed, this particular list
was developed for the hospital law statutes. Now ,
as in the Uniterm system, if you want to find where
particular words are found in the same documents,
why you just find the inter-sections and look them
up. You can do this manually, but a computer can
do it too.

Let's go to the next slide ( Slide D ) . Suppose that
you present the computer with this question ; "Please
supply me with a list of the statutes, and copies
of the statutes that deal with the restrictions upon
the remuneration that officials may receive from

185

16.

2 .

Year

FEDERAL STATUTES IN FORCE IN THE U. S.

SLIDE B

Incorporation 57 130

153 154

360 383

Incorporators 147 ....

406 ...

355 ....

84 88

86 ....

10 11

132
155

138
156

146
157

Increase . .
Incumbered
Incurred . . .
Incurring . . .
Ind

12 96 153
279 280 357 358 359

SLIDE C

LAWYER'S QUESTION

Please supply me with a list of statutes and copies
of statutes dealing with restrictions upon remuneration
for officials of charitable corporations.

BASIC COMPUTER QUESTION

(a + b+c+d + e + f)

a = trustee
b=trustees
c— officer
d = officers
e=director
f=directors

(g + h+i + j + k+l)

g = compensation
h = compensations
i = salary
j = salaries
k=remuneration
l = wages

Imprisoned 59

Imprisonment 144

Improper 240

Improved 399

Improvement 87

Improvements 339

Improving 351

Inc 410

Incapable 102

Incapacity 102

Incident 146

Incidental 330

Include 77

351

Included 57

Includes 428

Including 11

242
352
428

Inclusion 337

Inclusive 1 70

Income 3

323
367
406

Incompetent 97

Inconsistent 196

Inconvenience. ... 2

Incorporated 46

356
397

186

232

227 386
389 393

148
351
140
376
383

20
245
354
429

345

339

368

407

216

129

366

403

431
175
412
404

26
250
391

47
344
374
409
116
244

136
367
407

260
424
431

37
274
393

133

348
376
415
118

246
368
413

262
431

110

279
394

134
350
384

421
122

249
369

SLIDE D

charitable corporations." Now, there are certain
basic concepts here. You have the related terms
trustee and trustees. You would have to put both
the singular and the plural into this type of ques-
tion, as the program has been developed so far.
Officer, officers director, directors; these are all syn-
onyms, they have the same concept in common.
Here is compensation, compensations-salary, sal-
aries, remuneration and wages. I do not know how.
many of you know how to read this kind of equa-
tion, but if you do not you had better learn. This
is symbolic logic. You must be able to communicate
not just with the computer, "er", but with the com-
puter, "or", and this is the language that he speaks.
What you want here are documents that include J
either the word a, or the word b, or the word c, I
or the word d, and so on; and the word g or the
word d, or the word h, or i, or j, or k, etc. This j
is a logical product of two logical sums. Let usj
see what the computer produced. Here is a list)
(Slide E) of the hospital law statutes of the United j
States, which deal with that subject as located in
the computer with this type of question. The docu-
ment numbers are on the right; the citations are |
on the left, the lawyers who saw the demonstrations i

Flo. Stat. Ann. = S 155.16 (1943) 001

Fla. Stat. Ann. = S 155.16 (1943) 007

III. Ann. Stat. C. 24 , = S 44-6 (Smith-Hurd

1942) 009

Ind. Ann. Stat =S 22-3229 (Repl. Vol. 1 950) 01 2

Iowa Code Ann. = S 347.16 (1949) 015

Mich. Stat. Ann. = S 14.1139 (Rev. Vol.

1956) 016

Mo. Ann. Stat. = S 205.270 (1952) 019

N.C. Gen. Stat. = S 131-19 (Recomp. Vol.

1952) 029

N.C. Gen. Stat. = S 1 31-28.1 7 (Recomp. Vol.

1952) 030

Wyo. Comp. Stat. Ann. = S 44-1009 (1945). . 145
Cal. Corp. Code Ann. = S 10206 (Deering

1953) 146

Neb. Rev. Stat.^S 21-1503 (Reissue 1954). . 147

Nev. Rev. Stat. = S 81.310 (1 957) 148

Nev. Rev. Stat. = S 85.050 (1957) 149

Wyo. Comp. Stat. Ann. = S 44-1009 (1945). . 150

Fla. Stat. Ann. = S 155.18 (1943) 183

Nev. Rev. Stat.^S 450.440 (1957) 205

Conn. Gen. Stat. = S 12-81 (1958) 338

Me. Rev. Stat. Ann. C. 91-A, = S 10 (Supp.

1957) 366

N.Y. Tax Law=S4 389

renn. Code Ann. = S 67-502 (1955) 415

Wis. Stat. Ann. = S 70.1 1 (1957, as Amended

Supp. 1959) 431

SLIDE E

of this method in Washington, D.C., in 1960, were
delighted to go home and find that the computer
selected the citations in their respective states.

Now, instead of getting out just a list of citations,
such as patent numbers, you could get out the full
text. This could be a portion of text that is perti-
nent to your subject; or could be full text. Let's
show the next slide ( Slide F ) . Here we see portions
of some of the statutes referred to in slide E.

Now, how many of you, have guessed that the
word patent occurs more than ten times in the patent
statutes? How many have guessed that it occurs
more than 100 times? Very good. The word "pat-
ent" does occur over one hundred times in the Penn-
sylvania Statutes. Here is the print that was made
last Saturday at Pittsburgh. I also asked John Horty
to print out citations of statutes which have the word
copyright, which have the word invent, or any other
word having that root, the word patent, with letters
patent, and so on, and trademark and servicemark.
Would you like to pass that around for those who
are interested?

Now, if you record full text you have not pre-
judged the questions which will be asked. If you
record full text you have not lost information. If
you abstract you will lose information, and you will

Neb. Rev. Stat. = S 21-1503 (Reissue 1954). . 147

=S 21-1503 Stock, Dividends, Salaries,
prohibited.

Such (nonprofit) corporation shall have no
capital stock, and shall pay no dividends or
salaries to its incorporators or board of directors.
Nev. Rev. Stat. = S 81.310 (1957) 148

= S 81.310 Powers of corporation.

The (nonprofit) corporation shall, as an inci-
dent of its purpose and without any necessity
for expressing the same in its articles of incor-
poration, have the following powers, which it
may exercise in full measure without the neces-
sity of obtaining any order of court by author-
ization, approval or confirmation.*

* * * (5) To appoint and pay officers and
agents to conduct and administer the affairs of
the corporation, but no member of the board
of trustees shall receive any compensation.
Nev. Rev. Stat. = S 85.050 (1 957) 1 49

=S 85.050 Trustees not to receive compen-
sation.* Exception: No trustees of the (non-
profit) corporation shall be entitled to any
compensation except under some special em-
ployment of the board, or authority expressed
in the original deed or instrument of trust.
Wyo. Comp. Stat. Ann. = S 44-1009 (1945). . 150

= S 44-1009 Officers receive no salaries.

No officer of any corporation formed under
the provisions of the . . . 9th sub-division of
= section 1 (=S 44-1001 =Hospitals and in-
firmaries) shall receive any salary or remunera-
tion from such corporation for his services as
such officer, . . .

SLIDE F

have to go back and make new abstracts from time
to time, as has been said here so many times.

This particular material, the Pennsylvania Stat-
utes, includes six million words. Half of them,
three million, are insignificant. They are articles,
prepositions, conjunctions, and so on. These are
insignificant from some viewpoints, at least if you
are searching by key words; three million of them.
The entire text is recorded on four reels of tape.
About twenty minutes to one hour is required to
make a single, complete search. As many as 500
terms may be utilized simultaneously in making
searches, and as many as ten searches can be made
simultaneously. However, in that case the search
time is increased.

Other material has been recorded too, such as
all of the Maryland Health Law Statutes, and other
statutes. The New York Statutes will be going into
the tapes soon, and so will the New Jersey Statutes,
and then at that time some comparative language
studies can be made.

187

Due to the shortness of time, I want to skip some
of this material and talk a little bit about patent
claims. I have some copies of a paper here that
I have written. ( 3 ) . This is a mathematical theory
of patent claims, and I am not going to go into it,
except to show some of the work that has been done
here. Let's first take a look at the next slide
(Slide G).

SLIDE G

1. An electric light including:

a base having two mutually insulated electrical
contacts mounted thereon,-

a transparent envelope rigidly connected to said
base,- and

a coiled metal filament within said envelope, the
two ends of said filament being electrically con-
nected to said contacts.

2. An electric light including:
a base

a transparent envelope sealed to said base;
two electrical conductors extending through said
base into the space enclosed by said envelope,- and
a coiled filament within said envelope.

3. An electric light including:

a base composed entirely of insulating material,-
two electrical conductors projecting through said
base along parallel paths,- and
a tungsten filament having its two ends connected
to the respective conductors on the same side of
said base.

4. An electric light including:

a base composed entirely of insulating material,

two electrical conductors projecting through said

base,-

a tungsten filament having its two ends connected

to the respective conductors,- and

an evacuated transparent envelope sealed to said

base and enclosing said filament and the adjacent

parts of said conductors.

SLIDE H

1. The object is an electric light. The light include;'
a base. Two mutually insulated electrical contact;!
are mounted on the base. The light includes c|
transparent envelope. The envelope is rigidly]
connected to the base. The light includes a filal
ment. The filament is composed of metal. The;
filament is coiled. The two ends of the filamen |
are connected to the contacts. The filament i; ;
within the envelope.

2. The object is an electric light. The light include;
a base. Two electrical conductors extend througl -
the base. The conductors extend into the spacs
enclosed by the envelope. The light includes c
transparent envelope. The envelope is sealed tc
the base. The light includes a filament. Th«
filament is coiled. The filament is within th«i
envelope.

SLIDE I

Suppose that we were getting a patent on this
electric light bulb a few years ago. Some claim;
might have been allowed, like those on this next*
slide ( Slide H ) . Note that they are written in tht
new format that is now approved by the Unitec
States Patent Office. Notice that the claims refei
to parts such as a base, an envelope and a filament
That is one type of element in a claim. The claim;
also refer to the properties of the parts such a&
transparency, coiling and tungsten composition
And then we also have relations such as insulating
enclosing and connecting. We have three types of
concepts involved here — parts, properties, and re-
lations. In American patent law, as you know,
claims may define intersecting classes. An inven-
tion may be covered by one claim or another, but
not both. This is not true with most foreign coun-
tries. Usually if an object is covered by a later
claim of a foreign patent, it is also covered by an
earlier claim of this patent. In effect you have one
broad claim and then narrow ones inside of it. But
under our system, claims in the same patent may
overlap.

Now, the first step, I think, in making claims
easier to search is to rewrite them. Let us take a
look at the next slide ( Slide I ) . Here, the context
of the two claims has been rewritten at strings ol
sentences. Each of these sentences must be true
about an object for it to be covered by one claim.
Each of these sentences must be true for the object
to be covered by claim two. Now, if we can plot
this kind of information, we can analyze a claim
problem with a computer, in a matrix. Let us look
at the next slide ( Slide J ) . Here are the twenty
sentences required to express the four claims of this

188

Manufac-

turer's

aims

require-

ments

The object is an electric light.

The light includes a base.

The base is composed entirely of insulating material.

Two mutually insulated electrical contacts are mounted on the base.

Two electrical conductors extend through the base.

The conductors extend into the space enclosed by the envelope.

The conductors extend along parallel paths.

The light includes a transparent envelope.

The envelope is rigidly connected to the base.

The envelope is sealed to the base.

The envelope is evacuated.

The light includes a filament.

(1)

The filament is composed of metal.

The filament is coiled.

The filament is composed of tungsten.

The two ends of the filament are electrically connected to the contacts.

The two ends of the filament are connected to the respective conductors.

The connections between the filament and the conductors are on the
same side of the base.

The envelope encloses the filament and the adjacent parts of the
conductors.

The filament is within the envelope.

SLIDE J

hypothetical patent. If there is a "one" opposite a
sentence, this means that that sentence must be true,
for the object to come under the claim in the cor-
responding column. Now you can write these
claims then, individually, as mathematical formulae.
You can also describe the prior art in the form of the
same sentences, using something like a semantic
code, such as that which has been developed at
Western Reserve. If you write your claims this
way and if you describe your prior art this way, then
you can do some automatic searching which will
anticipate individual claims, not just locate broad
ideas.

Now, in just a couple of minutes, I will tell you
about something I have done with this type of
analysis. Here is a list of manufacturers' require-
ments (column 1 of Slide J). Wherever there is
a plus sign opposite a sentence, that sentence must
be true for a company to be manufacturing electric
lights at all. And the question is, how can a manu-
facturer meet those requirements and avoid all the
claims. Let us take a look at the next slide (Slide

p=1*2*4*8*9*1 2*1 3*14*16*20

+ 1*2*5*6*8*10*12*14*20

+ 1*2*3*5*7*12*13*15*17*18

+ 1*2*3*5*8*10*11*12*13*15*17*19 (E3)

r=13*14+14+3*7*15 + 3*11*15 (E5)

r ' = (13'+14')14'(3'+7 / + 15 , )(3' + ir + 15') (E6)
r '=14'15'+3'14'+7'11'14' (E7)

7' The conductors that extend through the base are
not parallel.
11' The envelope is not evacuated.
14' The filament is not coiled.

SLIDE K

K) which is just one of a series that shows how to
do it. You use mathematics. Each claim is repre-
sented by a logical product of a string of sentences,
each of which must be true to meet the requirements.
If you remove the sentences that relate to the re-
quirements, you have a residue here (Equation E5 ) .
When you negate Equation 5, you end up with
three classes of objects which meet the requirements
of a manufacturer (Equation E7) but lie outside
the scope of the patent. Each of these classes can
be translated into plain English, which in the case

189

of the last class means: if you will manufacture
an electric light in which the conductors that extend
through the base are not parallel, and the envelope
is not evacuated, and the filament is not coiled, you
meet your manufacturing requirements; but your
light will lie outside the literal scope of the claims.

Now, if you do not like mathematics, you can
use a computer program that I developed to do the
same thing. Here is a printed output (Slide L)
that tells the President of a company in plain Eng-
lish what he must manufacture to avoid the claims.

With this system, you can take the string of sen-
tences, you can take descriptions of objects, or
descriptions in the prior art, you can match them
up, and you can do your searching automatically.
Some day I am sure that will be done.

Mr. John Doe, Attorney
10512 Maine Street
Hometown, U.S.A.

To avoid literal infringment of the claims of Patent
No. 156038, and still conform to the requirements of
your client Mr. Harry R. Smith, as set forth in your
letter dated September 10, 1961, your client should
manufacture devices which are described in one of the
following paragraphs,
either

The base should not be composed entirely of in-
sulating material,
and

The filament should not be coiled,
or

The conductors should not extend along parallel
paths,
and

The envelope should not be evacuated,
and

The filament should not be coiled,
or

The filament should not be coiled,
and

The filament should not be composed of tungsten.

The foregoing conclusions deal only with literal
infrignment. Problems involving the doctrine of
estoppel, the doctrine of equivalence, and other
legal doctrines often employed by the courts in the
interpretation and application of claims have not been
taken into account.

Respectfully submitted,

SLIDE L

Reed C. Lawlor.

Now, there is other work going on in the applica-
tion of information technology to laws. People are
interested in predicting court decisions. The tech-
niques are being developed actively, and they are
working real well. There are also people who are

190

devoting their time to the use of symbolic logic for J
the analysis of legal documents, decisions, statutes,
language, and so on.

A paper describing a method by which 40 legal
points were utilized to characterize fifteen cases, were i
presented at the 1961 St. Louis convention of the
A.B.A. (4).

To those of you who may be interested in follow- !
ing the work of the Electronic Data Retrieval Com- i
mittee of the American Bar Association, I would I
suggest that you subscribe to MULL. This is a J
unique magazine. It is devoted entirely to appli- '
cations of information technology to laws ( 5 ) .

In closing, I want to read a little poem, a para-
phrase of one written by Ted Lassagne, of Libra- j
scope.

This is the tale of the 1 40 1 ,

The patent searcher that was nobody's son.

It spends its days in a futile hunt.

For prior art and argument.

After it finds them, it burns with shame,

It's obvious, says Examiner, just the same.

BIBLIOGRAPHY

1. Ford, Corey, Guide to Thimking, pp. 38, 60, 78

(Doubleday, 1961).

2. Horty, John F., "Experience with the Applica-

tion of Electronic Data Processing Systems in
General Law" 60 M. U.L.L. 158; See also I
Proceedings of Electronic Data Retrieval Com-
mittee for 1960 (Bureau of National Affairs, j
$2.00). |

3. Lawlor, Reed C, "Analysis of Patent Claims by i

Mathematical Logic" 796/ Summary of Pro- j
ceedings (Section of Patent, Trademark and]
Copyright Law, American Bar Association,
$1.50).

4. "1961 Proceedings of Electronic Data Retrieval

Committee" (in Press, 1962 M. U. L.L.).

5. Subscriptions to M. U. L. L. may be obtained at

$4.00 per year from the American Bar Asso-
ciation, 1155 East 60 Street, Chicago 37, Illi-
nois.

Mr. Glaser: Our speaker has made such breathless
headway, that we have time for some questions, as !
well as for examination of the exhibits that he)
brought. You ought to tell them what MULL
stands for.

Mr. Lawlor: Well, MULL is an abbreviation fori
Modern Uses of Logic in Law. Things go so fast j

nowadays, and people just cannot keep up with
knowledge, that our Committee published this maga-
zine in violation of all of the rules of the American
Bar Association, and it took a year and a half to get
their approval for their endorsement.

Dr. Bigelow: Sir, how would you code in your lan-
guage the process which involves the three ingre-
dients for a color composition which becomes say,
semi-solid and turns green and shows stride?

Mr. Lawlor: Break it up into primitive sentences.
Say, three substances are mixed. First, second and
third. The first substance is so and so, the second
substance is so and so, and the third substance is so
and so. In other words use plain English.

Mr. Kirsch: The text that you show in the regu-
larized form of the claim, is what Harris of the Uni-
versity of Pennsylvania, technically calls a kernalized
text. I think it is only fair to point out that several
years of research have been devoted to attempting to
make precise, the process that can be used by peo-
ple in particular and hopefully, ultimately by ma-
chines, in order to, in some systematic way, produce
the kind of regularized claims that you show. Harris
has suggested and apparently he subscribes to this
thesis, that the kernels of the text are a very useful
surrogate of the text for retrieval purposes. How-
ever, whether or not one can do this systematically,
even if one is a linguist, possessed of the usual re-
sources that linguists have, is at the moment some-
what questionable. I think that in fairness ot those
people who are less familiar with the technology here,
we should point out that it is quite conceivable that
one might do what you are suggesting systemati-
cally, regularly or conceivably even automatically but
we are rather far from that at the moment.

Mr. Lawlor: I do not think so. Whenever experi-
enced patent attorneys write patents, most of them
write the claims first. They then write the specifica-
tion. They use the claims as a basis for writing sen-
tences. There is no reason why those sentences could
not form the claims. We do not have to wait for the
linguist to make progress. We use language every
day. We file patent applications every day. We
can do it now. We do not have to wait for theory.

Dr. de Haan: I think it was a very interesting speech.
We have seen much of the contents of the Penn-
sylvania laws brought in to a computer. But I feel
inclined to ask the speaker, who is himself a lawyer,
if all this data you get from the computer are of any

help for you as a lawyer, to give an interpretation
of the Pennsylvania law?

Mr. Lawlor: This is all experimental. There has
not been enough work done to be able to answer
that question. I am convinced though that it will
help. I plan to send this particular digest to one
of the patent law associations in Pennsylvania, and
suggest that they publish the list of Pennsylvania
Statutes that mention letters patent, for the benefit
of the patent attorneys. I am sure a few of them
have them.

Mr. Fairthorne: I would like to ask whether the
word logic here has been used really quite fairly?
A hundred and twenty years ago, a form of algebra
was developed for a very narrow type of what might
easily be called logic, although it certainly does not
carry the same weight as what logic does to most
people. Historically, that form of symbol manipula-
tion that was used has all characters of the word logic
in it, but no longer necessarily has any connection
with logic at all. You go to Boolean algebra or
something, and it is most unfortunate that people
think that Boolean algebra is symbolic logic. Now
under certain circumstances, it is valid to use it in
logical situations. I am not here at this moment
making any criticism one way or the other, although
I certainly will in the future, on this matter. But
do be careful, because we are dealing with a form
of symbol manipulation. Dr. Bigelow, I think,
has mentioned one thing that cannot in fact be put
in terms of the binary function which are binary
operations. You can not get ternary relations into
binary except the most exceptional, and so on. One
has to be very careful.

And the other thing is, what entities do these
letters stand for? Do they stand for concepts? Do
they stand for word occurrences? Do they stand for
documents? The algebra is perfectly valid for sets
of objects. Doubtless to say you can have products
of sets, things that are commonly two sets and so
on. But there is no guarantee ab initio, that a set of
words has a corresponding set of documents that is
of the slightest use to you, or that the set of words,
or the set of concepts, or the set of documents in exact
correspondency work. I am not saying they are not
there. But I say you must not assume that they are.
And I do think that there are a great many things
that need to be made a little clearer, before we can
discuss details. My own views on this matter are
that Boolean algebra is grossly exaggerated unless

191

it is very carefully handled, but that is not at the
moment important.

Mr. Lawlor: I would like to add one comment to
that. Boolean algebra is only a very elementary
form of symbolic logic.

Mr. 'Glaser: I believe we still have a chance for one
more question before we will be deprived of our
lunch, which would be a very serious matter.

Mr. Kirsch: I think the crucial issue here, and one
which I would like this afternoon if I may, Mr.
Lawlor, to elaborate on, is this. Shall we present
to the community of people who are concerned with
solving some eminently practical and difficult prob-
lems, a class of techniques which will be rather
costly for them to become familiar with, and to sug-
gest to them, which I would suggest is inappropriate
at the present time, that these techniques offer solu-
tions to problems rather quickly? Shall we present
these techniques to them in this guise and incur the
possible danger of misunderstanding and a con-
sequent disillusionment? Or shall we attempt to
present these techniques as techniques which have
some very limited, conceivably ambitious goals, but
not attempt to oversell them and consequently rep-
resent appropriately, the time scale, the cost, the
amount of energy that is necessary in order to achieve
the kinds of things that presumably you and I are
both interested in being able to do? That seems to
me to be the crucial issue here.

Mr. Glaser: I think I will have to, to stay on our
schedule, end this part of the meeting. The course
of a working group is such that we are back to our
original schedule, having covered exactly the things
that are on the printed program. We will move
Mary Stevens brief talk on character recognition
and pattern reading, to the afternoon. Therefore,
the afternoon will consist of the presentation of
Russell Kirsch, the one of Dr. Heilprin and a brief
talk by Mary Stevens. Then as I announced this
morning, Dr. deHaan, and a discussion of Dr.
deHaan's suggestions after that. That will con-
clude the afternoon session and the conference.

A recess for lunch was called by the Moderator
and the meeting reconvened at 2:00 P.M.

AFTERNOON

Mr. Glaser: We have a very ambitious program, so
we want to start promptly. I should first like to
explain that both the Commissioner, Mr. Ladd, and
the Assistant Commissioner, Mr. Fay, of the United

192

States Patent Office, were forced to be away this
morning, because they were in a hearing before the
Budget Bureau. I am sure you all know enough
about the American System of government to know
that that has something to do with money. We do
not like to seem too mercenary in the Federal Civil
Service, but the fact is that these matters must be
given a high priority. Mr. Fay is back with us now
and will have something to say in a few moments.

This group is not a duly constituted parliamentary
group, it has no charter and therefore can pass no
resolutions. I will ignore the niceties of this legal
situation, and entertain two resolutions, if the group
is so minded. I think I would be willing to con-
sider first a resolution on the part of the group,
thanking our hosts for the hospitality they have of-
fered us, partly in the form of lunches so large that
we will have trouble staying awake for the afternoon
session. Our hosts are of course, the United States
Patent Office, the Patent Office Society, and the
National Science Foundation. In order to sup-
press debate on this matter, I would be willing to
entertain a resolution which merely states the sense
of the group and does not require a detailed draft-
ing job. — We have a motion. In the absence of
opposition, I will instruct whoever it is should do
these things to communicate with the three
sponsors and thank them for their colorful and
generous hospitality.

I will also, now that I have a precedent, entertain
a second resolution, this one to the program com-
mittee (and since I am not on any of these things,
by the way, I can do this in good form) which was
headed by Harold Pfeffer and Mr. Gorecki of the
Patent Office, for their work in organizing this
Workshop, which seems to have had just about the
right amount of scheduled program, and time for
us to talk about the things that really concern our-
selves. I will therefore, entertain a motion to thank
them for their hard work on this conference. — In
the absence of debate, I will assume according to
our unwritten charter that these resolutions have
been adopted unanimously.

And we now will proceed with the program. The
program has been rearranged yet again. Dr. de
Haan's presence here is sufficiently important that
everybody in town seems to want him for something.
Therefore, we are going to move at once to the part
of the program which was originally labeled "Con-
cluding Remarks." I already warned you we are
not going to be concluding at all but merely starting.

After his remarks, Commissioner Fay of the
United States Patent Office, will say a few words
and I think they will have something for us to dis-
cuss. This will carry us up to the coffee break.
It seems a shame to talk about food so soon. And
then we will proceed with the papers that are on the
program. Without further delay then, I will turn
the meeting over to Dr. de Haan.

Dr. de Haan: Thank you very much, Mr. Glaser.
Ladies and gentlemen, first of all, I have to apolo-
gize to the speakers who are scheduled in the pro-
gram at this moment, for the fact that I have been
brought in before them. My indiscretion is only
due to the fact that I have been invited by Com-
missioner Ladd to say something in the Patent
Office, to the examiners of the Patent Office, about
the Netherlands legislation on patents at four o'clock.
That would normally be the time that was reserved
for me for giving my concluding report and remarks.
I am very sorry that at this moment I am not in a
position to also consider in these remarks the con-
tents of the very interesting addresses which will be
made after I have left this meeting. Ladies and
gentlemen, I was a bit surprised when I was brought
into this program to give the concluding remarks
of our three days meeting here. Personally, I
thought that a U.S. attendant of this Workshop
was much more qualified than I, myself, to give con-
cluding remarks. I apologize once more that I have
to speak in a language which is not my own, and
even as the situation requires, that I have to speak
about a matter which is not as familiar to me as
the Netherlands Patent Law.

I shall try to do my best, and give some observa-
tions about what we have discussed, my impressions
about it; and then I shall try to come to conclu-
sions. We started Monday morning, after the very
interesting introductory remarks of Commissioner
Ladd and Commissioner Watson, to attend lectures
about the existing classification system. We heard
about the International Patent classification from
Mr. Rubach, the United States patent classification
from Mr. Vincent and the British Classification Sys-
tem from Mr. Knight.

Allow me to make a very general remark about
classification systems. What is the aim of all classi-
fication systems used by examiners, either Patent
Office examiners, research examiners or industrial
examiners who have at their disposal tremendous
files of documentation? Well, the aim of classifi-
cation is to bring together the documents which are

related to one another, so that the examiner does
not have to screen the contents of all his documenta-
tion. That way he knows where he has to examine
and which may be relevant documents for the search
he has to make, and he can leave aside those docu-
ments which according to the classification system
are of no interest to the question with which he
has to deal.

Now, we have the existing classification systems,
which the speakers explained to us Monday. Well,
I have already observed that in my opinion, all three
classification systems are as good as they can be and
that any preference for any one of them is, in my
opinion, not very well founded. I think an exam-
iner accustomed to it can work with it. But these
and other classification systems (except perhaps the
British System) all have in common that documents
are introduced into the system after usually one
entry, perhaps with some superior entries, but each
document gets a place into the system. The differ-
ences of the existing classification systems are that
the entries are different, so that for instance, in the
U.S. classification system and the IPC and the Ger-
man and the Netherlands Systems, the individual
documents are classified in quite different groups
and subgroups and in the patent offices they are
placed in different pigeon holes.

Now the difficulty with the existing classification
systems is, if I may speak now only about the patent
offices, that the inventors are so very unkind as not
to take into account the classification systems with
which we work. I mean to say that each patent
application, and also each granted patent, contains
some subject matter which according to the existing
system can be classified in more than one class or
group, and pigeon hole. But as we have so many
specifications and so many documents that we are
unable to classify them in many groups, it must
suffice to put them only in the groups and pigeon
holes which seem to our classifiers the most impor-
tant ones. And that is the reason that when
searching in a certain classification system, it is
very difficult to know if in searching in one class
or group, you have really found all the relevant
matter. We are all accustomed to the type of
patent application that requires a search in several
groups if one desires a thorough examination.

To come back once again to the different systems,
we also discussed the possibility of a nixing at a
uniform classification system, on what I might call
the old basis as we use it now. I have observed
that I think this would be very difficult. I men-

193

tioned that our existing classification systems are
heavily burdened with tradition based upon ex-
perience and that it would be difficult for the various
patent offices to break this tradition. There is still
another reason. You must realize what a tremen-
dous amount of capital is invested in the existing
classification systems. If we should like to arrive at
a new system or to a coordinated system, it would
be necessary to reorganize our entire documentation
system. That can be done, but only very slowly,
and only where it is necessary that some class be
reorganized. Then you can perhaps reorganize
into another system as we intend to do in Europe,
where there is not too great a difference among
classification systems in the countries which have
accepted the International Patent Classification.
But I think that at the present moment this calls for
our discretion. There is a much more important
matter, viz., what we call the mechanized search,
and I think it has been made sufficiently clear
during these days, that if we desire to come to a
mechanized searching system, a complete revolution
of our classification and documentation is necessary,
and this revolution will take place in all the patent
offices who want to change over to a mechanized
system. It is an advantage, that we are now all on
the same level. I mean we have to start something
new. And if we start something new, international
cooperation in this field is much easier than if we
have to bend existing systems to a common end.
And so, I should like to speak now about the general
topic of our Workshop, the mechanized system.

Quite a lot has been spoken about terminology,
and first of all I should like to trace what we under-
stand of the mechanized system. It is a system in
which we perhaps use machines, but I do not think
that that is the most relevant characterization of
the system. What we in fact understand (and I
hope to hear your opinion about this) of a mecha-
nized system, is that we introduce subsidiary filtering
means between the searcher, the examiner, and his
files. Thereby it is possible to give to the examiner,
on his desk, only the relevant documents and prefer-
ably more of them than he can get with the existing
searching systems, in which he makes his search
among the documents themselves and has to study
quite a lot of them, of which he finds the greater
part to be of no interest at all. What we want to
achieve with a mechanized system, is not so much
to prevent an examiner from studying documents,
as to have the means to give the examiner on his
desk less, but more relevant documents than he can

get at the present time. And to achieve this end, we
want an intercollation of some system, some tools i
which help us to select the relevant documents. <
These tools, Mr. van Wassbergen said this morning, j
may be lists in which the relevant features of each I
document are in a certain order, written down so
that the examiner can determine which documents
are important. It may be a punched card for man-
ual selection. It may also be a punched card for a
machine. And in the last instance, it may be a j
magnetic tape of a computer. Well, if we take the
list and the simple peek-a-boo cards we see that j
these can be used without any machine. The
punched card for a business machine, the magnetic II
tape for a computer, it is self-evident must be han-
dled by a machine.

Now, to avoid all misunderstanding, I should like
to say that it is perhaps a bit premature to aim at a j
fully mechanized system. I should prefer that |
we aim at a system in which we may eventually j
use machines. I think everybody agrees with me I
that a manufacturer, who uses machines in his plant,
does not yet have a mechanized production system. ;
Therefore, it is quite important and I should like .
to draw the attention of the linguists to it, that in !
our discussion, we give a definite, specific meaning ;
to the word "mechanization" and that we should |
not use it in the sense in which it is used in industry. J

And now about these tools which are put between 9
the examiner ,and his files. Well, they can be of all J
kinds. We have heard very interesting talks about jj
abstracts from Mr. Jeffery of Australia, from Mr. jj
Kotaro Otani of Japan, and from Mr. Wolk. Even \\
so, these are tools which are generally not for ma- j
chine searching systems. But abstracts are made to i
facilitate the retrieval of relevant documents for a j
search. And we have heard that many searches J
can be done very well with the help of abstracts. J
I think that to a certain extent also, the holes I
punched in a card or the symbols on magnetic tape .
are abstracts. They all contain the most important I
fractions of a document in such a manner that we i
can use them to retrieve this document. What now j
is our task? It seems to be that we investigate those |
tools that give us the possibility to retrieve j
documents. j

I think that the result of our discussion in the past
days, is that depending on the art in which the
search is made, special kinds of tools can best be
used. In many parts of the art, it will be sufficient
to work with lists or with peek-a-boo cards. In
more complicated fields of technique, we have to

194

change over to punched cards for business machines
and perhaps there are arts in which we are obliged
to use a computer.

I feel our first task, if we desire to arrive at what
you call a mechanized system, is to consider what
tools we should use, how we can make them, and
what system we want in order to retrieve relevant
documents.

There was also a discussion about preparing these
tools automatically. The automatic preparation of
abstracts, the automatic transcription of documents
onto a punched card, and so on. I think it is very
interesting and I hope that those who are now doing
research in this topic will succeed as soon as possible
in finding a method for automatically abstracting
documents. But at the present moment, we do not
yet have a system at our disposal which would be
sufficiently elaborate to satisfy us Patent Office men.

So I think as we all know, that our Patent Offices
are in very great need, and that without sufficient
manpower we cannot go on with our hand search-
ing. We must find some way to work toward what
we call a mechanization. And so I think our first
task is to find means to establish those tools which
serve for retrieval of relevant documents. These
tools differ, as I have said, from system to system.
And so I think we must first of all consider the
various arts and which system would be most ade-
quate for each art as has already been done for some
arts in the Australian and the British and Interna-
tional Patent Institutes. If we do that, I think the
use of the system of lists and the system of peek-a-
boo cards will not give too much trouble. But if we
want either business machines or computers, then
I think we have to make a very thorough study of
how to organize the information in documents in
such a way that put into the machine and handled
by the machine, they will give us not a direct answer
about the novelty of an invention, but will give us
the most relevant documents which have to be
studied in order to judge the novelty and patent-
ability of such an invention.

And I feel that this work should be done in inter-
national cooperation. We all know, if we want to
make such tools for each document, we have to
analyze these documents. I heard from Mr. Pfeffer
yesterday that he thinks the analysis of one docu-
ment, takes about five to six hours. Suppose the
U.S. Patent Office has in its files six million of docu-
ments, the analysis of all the documents would take
thirty million man hours, which means twenty
thousand man years. In other words to analyze

the entire documentation of the U.S. Patent Office,
would lead to the fact that the Patent Office should
be closed for twenty years and that its entire staff
would do nothing else but analyze documents. You
see, it is a tremendous amount of work and I hope
that the six hours mentioned by Mr. Pfeffer can be
reduced at least for the greater part of the arts
involved.

But all the same, we all know that the U.S. Pat-
ent Office is by far the largest office of the world,
and I myself, who am in charge of a small patent
office, am sure, as you can very well understand, that
it is quite impossible to fulfill such a task with my
own means. And there are many other patent of-
fices which are in the same position. I think that
for that reason, it is quite necessary that we col-
laborate, at least if we wish to achieve our ends as
soon as possible.

And in what way shall we collaborate? I have
quite a few ideas. I do not know if these ideas would
be agreeable to you and I do not even know if these
ideas are agreeable to my own Government, but I
thought that perhaps an International Committee
might be constituted to work out a research program.
Just as in many industries there is a board which
is charged with making plans for research, in which
all relevant questions can be dealt with, there could
be considered for each art, which system is the best,
and who should design a search system and test it
and elaborate some plans as has been done already.
That can all be coordinated. And it seems to me
that this work of actual researching could be taken
over by the several Patent Offices which are inclined
to cooperate in this field.

I think that in this field if we do not come to a
real international cooperation, perhaps the U.S. Pat-
ent Office may succeed on its own. But as to the
small European patent offices, I seriously doubt that
they will ever come to a result, and Mr. Moderator,
I should appreciate it if you would be kind enough
now to discuss the possibilities of such an interna-
tional committee which, after establishing a re-
search program might be followed by further
research in coordination between several patent of-
fices. Thank you very much.

Mr. Glaser: Thank you very much. Dr. de Haan. I
believe the Assistant Commissioner, Mr. Fay, has a
few words he wants to say before he has to rush
off to somewhere else. Then we will throw the floor
open for the discussion of the suggestion we just had
from Dr. de Haan.

195

Chairman Fay: Thank you, Mr. Glaser. We do
not have to rush off that fast. Dr. de Haan has said
he would like to listen to the discussion about his
proposal, so that we will stay until the coffee break.

First, I want to introduce to you some people from
the Patent Office whom you may not have met.
First Assistant Commissioner, Edwin Reynolds, Mr.
P. J. Federico, Examiner in Chief, Mr. George
Roeming of the Solicitor's Office.

I think we can all agree that this workshop has
been worthwhile in one way or another to most of
us. While some other work assigned to me, did not
allow me to be here regularly, I have observed a gen-
uine enthusiasm for the work we are doing here,
among the various representatives present. I know
that we have learned from each other, and I trust
that what we have learned will be useful to each of
us in helping us to approach our own information re-
trieval problems more intelligently. I would like to
add my special thanks and appreciation to the speak-
ers who have led our discussion groups and who have
presented the material upon which we have based
our discussions. I think we can all readily agree that
the presentations were both stimulating and en-
lightening. Finally, before closing, I would like to
voice the hope that we will have many more of these
Workshop sessions in the future Mr. Knight of the
United Kingdom, in his presentation on Monday
afternoon, quoted to us from a report which stated
that these workshop sessions themselves are helpful in
solving our information retrieval problems. I agree
with that statement and I would add that the in-
formation gathered in the workshops, such as this
will prove only as valuable as the use to which we
put it. I trust that we will have at least, begun to
understand our information retrieval problems
through discussions here, and that we may have fu-
ture opportunities to meet and continue this work.

Mr. Glaser: Yes, Dr. de Haan.

Dr. de Haan: Mr. Moderator, I just spoke about
abstracting. I should like now to talk about
thoughts that have not been written down or said.
I mean, the thoughts, I am quite certain are now
in the minds of your foreign guests. We have all
greatly appreciated this invitation to attend the
Workshop. We have especially appreciated the
way in which these discussions took place; the open
and unprejudiced way in which you Americans face
the existing problems and the clear and very matter-
of-fact manner in which you deal with them. I
think it has been for all of us a revelation to be here,

and to cooperate in this very nice atmosphere of
friendship.

And then I want to thank you, too, for your real
generous hospitality on all occasions. Thank you
very much.

Mr. Glaser: Thank you, Dr. de Haan. I shall see
that your words are communicated to those who
deserve them. Now I take it we are still a parlia-
mentary, unparliamentary organization without a
charter, now about to discuss the suggestion that
Dr. de Haan made for the formation of some sort
of international cooperation committee, in order
that we not lose the advantage of having discovered
who each of us is and what our interests might be.
I would, therefore, like to hear from those who are
attending, especially the representatives of countries
other than the United States, initially, to see what
they think ought to be set up, what degree of for-
mality, how it might be organized, and so on. And,
of course, also to address themselves to the question
of whether or not there should be such a group,
which is, of course, the proposal that Dr. de Haan
has made. I then throw the floor open for free
discussion of this suggestion. I suppose it would be
appropriate to ask first of all whether anyone thinks
this is a poor idea and would like to speak against
it. This is the easiest way to get an argument going.
If no one rises to this, I shall assume that everyone
present thinks that the basic suggestion is a good
one. I will entertain any comments anyone cares
to make about how it ought to be organized and
how it might operate.

Mr. Jancin: I think it is a very good idea. How-
ever, I think that you almost need a committee to
formulate the committee. In other words, this is
something that probably should be discussed by a
nucleus of people to develop a working international
committee.

Mr. Glaser: A very sound suggestion. I do not
want to presume on the usual privilege of the Chair-
man to appoint committees because my position
here is already so tenuous. I do not know quite on
what authority I have this chair, and we ought to
proceed that way. However, if there are any com-
ments to be made here in the open meeting, this is
the right time to make them. If not, we will simply
have to discuss as informally as we can, the arrange-
ments under which we might set up such a commit-
tee and see that we maintain the lists of the
identification of the people who are interested in
this work, and who occupy positions of responsi-

196

bility in the various countries, and that we exchange
documents, and so on. However, aside from the
fairly obvious things that such a committee would
have to do, if anyone has any suggestions about how
it ought to be set up, how it ought to be organized,
who ought to be on it, or anything else, the floor
is open for that. Yes, Harold.

Mr. Pjeffer: I would like to suggest that the visitors
who are here today consider themselves as consti-
tuting an ad hoc committee to further discuss this
matter.

Mr. Glaser: Do we have any comment on this sug-
gestion? If not, we already have a committee.

Mr. Jancin: I would like to go a step further and
suggest that all of us give this some thought and
possibly direct our remarks to a single person that
you, Mr. Glaser, as Chairman, might designate at
this time.

Mr. Glaser: Yes. Let me suggest that Harold Pfef-
fer, who is in charge of the Arrangements Commit-
tee of this conference, be the person you can address.
This gives us at least someplace to which we can
send our remarks. If you simply send them to Har-
old Pfeffer, United States Patent Office, I am sure
he would get it. Meanwhile, I suspect that before
any of you get too far away, we will be a little bit
further along on the organization of such an ad
hoc committee. Is there anything further? Yes sir.

Mr. Rockman: I think one of the first things we
should do is to get the committee together to find
out what actually has been done in each of the
various countries — Australia with her 26 different
arts, the United Kingdom with their system, and find
out in each country which art actually has been de-
veloped, to what degree, and the means used to get a
set of descriptors. I think that would be a start
to combine all this actual knowledge we have now,
use that for a springboard and go on from there.

Mr. Glaser: If I may apply the rules of legislative
construction to what you said and take everything
literally, I do not think it is practical for the com-
mittee to get together and say in detail what each
of them are doing, since this would occupy a great
deal of time. I think, however, the committee might
very well arrange to accumulate these things in some
useful form so that they can be sent around.

Mr. van Waasbergen: I think, Mr. Chairman, that
these days have made it clearer that many people

do not know what other people are doing. There
is an exception for what you in the United States
are doing because you have, in a magnificant way,
published about 20 reports, which were available,
I think, to everyone else. But we at The Hague,
for instance, know very little about what is being
done in Stockholm, even in London, and Munich,
towns which are not so far from us. I think one of
the first things which we should do is establish some
means of comunications from all interested people
to all other interested people.

Mr. Glaser: This suggestion has so much merit I will
not even open it for discussion, but simply regard
it as one of the instructions to the ad hoc committee,
if that satisfies everyone.

Mr. Roller: I should like to suggest that the ad hoc
committee and the temporarily designated secretary,
that is the representatives of all of the visiting patent
offices, with Mr. Pfeffer, get together and try to find
some time to meet either in the rest of the day today
or sometime tomorrow and try to arrange either
for the establishment of a temporary secretariat for
the international committee or some point at which
some next meeting or some next formal communica-
tion could be had. I think it would be well for
them to get started at least in this formal way before
leaving these grounds.

Mr. Glaser: Again, I think this has sufficient merit
so it should be regarded more as an instruction than
a question for debate. Mr. Fay, did you have some-
thing you wanted to say?

Chairman Fay: Mr. Glaser, on behalf of Commis-
sioner Ladd, I think we should be glad to act, if this
group should so ask us to, to appoint a tentative or a
temporary chairman for a committee and see if we
could not, at least, begin the organization of some-
thing which could be considered at a subsequent
meeting. We could prepare policy, through the
committee, which would suggest a structure that
might be available for a subsequent workshop. I
think at this time in this workshop, it is a little late
to put together a committee that can come up with
very much policy now. We could do this in the
months to come. We would be glad to centralize
any information you would like and to assume to
appoint a chairman temporarily, if the consensus of
the group here would like us to.

Mr. Glaser: What is the wish of the participants from
the other countries with regard to this suggestion.

197

Mr. van Waasbergen: Mr. Chairman, would it be
perhaps a good thing to ask which people will be in
Washington tomorrow afternoon.

Mr. Glaser: Yes. How many of the delegates to
this organization will still be in Washington to-
morrow afternoon? We will have a show of hands.
Well, not all, but quite a few will be here. Do you
want to say anything about this tentative arrange-
ment, Harold?

Mr. Pfeffer: All the representatives will be present
for the meeting tomorrow with the Commissioners.
Perhaps sometime, either before or after the meet-
ing, we may be able to consider the organizational
problem.

Mr. Glaser: I think the first step will be taken then
at the meeting tomorrow with the Commissioner. I
take it that the American Patent Office proposal vol-
unteered a certain amount of secretarial service (this
being not a trivial thing in the organizing of a new
effort) somebody to keep mailing lists, to see that
things are reproduced that should be, and so on. I
take it you are speaking for the Commissioner in
offering this, Mr. Fay.

Chairman Fay: I think so. I think we can discuss
this proposal tomorrow, after the other subjects are
terminated.

Mr. Glaser: Fine. Well I think that does give us
a way to move ahead. Does anyone have anything
further by way of instructions to the committee or
comments on the proposal that might be brought
up at this meeting while we are all here? If not,
I think that concludes this part of the program. We
are a little bit early for the coffee break, which is
good because we have three papers yet before the
conference, which will be read in the second session
this afternoon and will complete the conference.
Needless to say, it is perfectly proper during coffee
break to corner people and tell them what you think
about this organization. That is what coffee breaks
are really for.

A short recess was called by the Moderator.

Mr. Glaser: We are ready to resume for the last
session. We have three papers of substantially dif-
ferent content remaining yet in the conference. In
all three cases I know enough about the speakers to
state that they could easily spend a long and profit-
able session on the topics that they will cover. In
the merciless manner in which we must proceed, all
three will summarize very briefly the kinds of re-

search they are conducting and the results which
they are obtaining. We will have first my colleague
Russell Kirsch of the National Bureau of Standards,
who is working on an approach to the Patent Office
problem through formal linguistics. He will make
his presentations now — "Theoretical Research on
Information Processing Potentially Related to the In-
formation Retrieval Problem."

Mr. Kirsch: I would like to paint a fairly black
picture with one little ray of light in it, and suggest
to you that it is largely your option to determine
how much can be extracted from this little ray of
light. The thesis that I would like to develop is
that there is a good deal of work of a theoretical
nature which although it is not immediately con-
cerned with the information retrieval problem, can
potentially be made to relate to the information re-
trieval problem. What I would like to address
myself to, then is a consideration of existing research,
predominantly of a theoretical nature, that, at your
option, can perhaps be made germane and useful
in understanding the information retrieval problem.

This is at your option for the following reasons.
The kind of research that I would like to talk about
is, in fact, going on at present and there is every
reason to believe it will continue to be conducted in
as effective, scientifically satisfactory a way as it is
at present. There is no a priori reason to believe
that this work will direct itself to the kind of prob-
lem that we have been discussing here unless some-
body fnakes a significant attempt to show how it
relates and to use whatever results are available.
This requires at least two very difficult decisions on i
the part of those of you who believe that some of
the things that I am going to talk about are related,
or may be made to be related, to the information
retrieval problem.

The first decision, which would have to be made
in the affirmative, is whether or not to engage in
the necessary training to understand the predom-
inantly mathematical or linguistic or computational
techniques that underly these particular theoretical
disciplines. !

The second decision which would also have to be
made in the affirmative by the members of the inter-
national patent communities, is to attempt to ab- j
stract from your problems those parts which are \
fundamental, systematic, not chimerical, but con-
sistent and repetitive from one system to another and i
from one time to another. Then you must com- |
municate this abstraction of your problems to the

198

members of the scientific community in such a way
that they can proceed to do something about it.
That is also a very difficult task, but some of us
have been addressing ourselves to it for years, with
some limited success.

Both of these decisions, if made in the affirmative,
are reasonably costly in time, in manpower and
patience. I would like to suggest to you simply
that the decision should be made without attempt-
ing to argue in which direction these decisions
should be made. My purpose here is to present to
you a description of the kind of research being con-
ducted so that you will have a little more informa-
tion on which to base your decisions.

First of all, there are many different versions of
the information retrieval problem. I have identified
a few in the past discussions. The version of the in-
formation retrieval problem that one works on is
determined partly by the problem that one has. I
have been very much impressed here by the great
degree of similarity that there is in the patent office
problems in one community and the next. But more
importantly, the version of information retrieval that
you formulate and attempt to work on is dependent
upon your idea of what intellectual and mechanical
tools are at your disposal.

I would like to trace very briefly the history of
kinds of information retrieval problems that people
have worked on with respect to the tools that were
presumed to be available to them. The earliest kind
of information retrieval problem was implementable
with hand-sorted cards — the use of the set inclusion
kind of information retrieval with descriptors is
easily implemented in this type of mechanical ve-
hicle. Consequently many people construe and
some still do construe the retrieval problem — that
is, their version — to be that kind which is im-
plementable with such devices.

A somewhat more ambitious version of the in-
formation retrieval problem presumed implementa-
tion with punched card machines. A still further
and perhaps more ambitious version of the informa-
tion retrieval problem presumes to make use of the
calculating power of electronic computers and here
|we find such versions of the retrieval problem as
statistical analysis of texts for the purposes of auto-
abstracting or perhaps it should be called auto-
textracting.

There have been not only mechanical tools avail-
able, but also formidable intellectual tools. For ex-
ample, many people in the past few days have been
talking about the use of classification theory. There

is a large literature on this theory. I just mention
in passing the use of statistical analysis of a sophis-
ticated kind, easily implementable on electronic
computers or even on some punched card machines.

I would like to continue the survey beyond here
and ignoring those things that I have just men-
tioned, tell you something about other tools. My
own particular familiarity is mostly with electronic
computers. In trying to use electronic computers
over a period of years for various problems, in par-
ticular the information retrieval problem, it has been
my own personal experience to discover that we
know very little about these machines and very little
about their use. Consequently we must develop
considerably more insight into these machines and
their methods of use and into the methods of char-
acterizing the information that they process in order
to attempt to do anything of a significant theoretical
nature with respect to the information retrieval
problem.

Let me talk a bit about some of these tools, both
mechanical and intellectual, which are at present
available and which you might perhaps be able to
apply to the information retrieval problem. I should
re-emphasize that none of the tools I am going to
mention have been used in any significant practical
way to solve any parts of the information retrieval
problem. That is not coincidental. I selected just
those tools which have not yet been made practical
use of, because I presume that you gentlemen are
eminently aware of practically available machines
and practically available tools. So let us discuss
some of the more esoteric tools.

First of all, there exist today fairly sophisticated
programming languages. The programming lan-
guage is an artifice invented by the computer pro-
gramming profession to make its job easier. People
have been writing instructions for computers for over
a decade now, and many people who have been do-
ing this were not only disturbed by the difficulty of
their task, but ingenious enough to use the very ma-
chines to which they were addressing their instruc-
tions as vehicles for making their own jobs simpler.
As a result of the kind of investigations that have con-
sequently grown up in the computer programming
profession, there exist today very powerful, formal
languages, artificial languages, invented predom-
inantly by computer programmers, which have a very
powerful descriptive capability of describing
computer processing operations.

More recently, programming languages have been
developed which have a more general area of appli-

645803 0-63-9

199

cability — that of manipulating general arrays of
symbolic information, not necessarily numerical and
not necessarily for the purpose of instructing com-
puters. So we have today in existence such things
as the symbol manipulations languages called LISP
and IPL-V and a few others, developed in various
places. A rather interesting one which is in the
early stages of development right now is the so called
"Advice-Taker" of McCarthy and Minsky at MIT.
The purpose of this language is to facilitate the
possibility that the operator of the machine might
give general advice to the machine about a particu-
lar task, which consequently would enable the ma-
chine to perform the kind of inference necessary to
perform very elementary tasks of a non-numerical
nature.

Another kind of tool of a more mechanical type
that has already been implemented is the class of
picture processing techniques that some of us have
been working at. We do have in existence today
the ability to feed diagrammatic or pictorial informa-
tion directly into machines. That is, there exist
devices which can scan an actual picture and copy
that picture in a rote fashion into a machine. The
possibility then opens up, and it is at the moment not
much more than a possibility, that one might write
instructions for such a machine to enable it to make
this rote-copy photograph and characterize it by do-
ing such things, for example, as counting the num-
ber of resistor symbols or finding out whether there
is a parallel tuned resonant circuit in the plate cir-
cuit of a vacuum tube oscillator in electrical circuit
diagrams. This has not yet been done.

Notice, I am mentioning this not as just a remote
possibility. The fact that I mention it indicates that
there are some people who are trying to achieve it.

Moving over to the more intellectual tools that are
available — those which have not been implemented
in any form of hardware, we have the whole tradi-
tion of Descriptive Linguistics. The linguistics com-
munity has attempted for a long time to find out
what is systematic about language, and it is taken
as a basic predicate in the linguistic community that
language is a very systematic phenomenon — a
phenomenon that can be studied the way any other
physical phenomenon can be studied. It happens,
unfortunately, to be a very complex one, but we
now have the ability to describe in particular to
machines, very complex phenomena and the ability
to exchange these descriptions through the artifices
of programming languages.

200

There are now investigations in descriptive lin-
guistics which are particularly concerned with such
things as automatically characterizing the syntacti-
cal structure of, say, English sentences. That is to
say, given a set of words, there are in general, very
few combinations of those words that form sentences
in any natural language. The vast majority of them
will not, and the attempt to elucidate these con-
straints and to exploit them, in particular for such
things as information retrieval, is the job of the syn-
tactical analyst. There exist computer programs
now which can do very elementary kinds of syn-
tactical analysis on natural language text. It turns
out that the size of the task is such that the number
of rules one might need to describe the structure of
a language, like English, could possibly be compara-
ble to the number of words in the language itself.
That is a rather sobering discovery if it is true.

More than passing mention should be given to
the subject of Mathematical and, in particular,
Symbolic Logic. Two things come to mind here in
the way of investigations that have some potential
applicability to the problem at hand. The logicians,
and more recently the computer users have been|
concerned for some time with the following rather
fundamental question : what can and what cannot,;
in principle, be calculated by machines. It turns
out that this is not a very new question. Its earliest
version in the recent literature dates back to the
early 1930's, and about something over a half dozen
investigators have independently attempted to char-
acterize this problem and to propose mechanisms:
which they thought had the general power to cal-'S
culate all that can be calculated. These mecha-|
nisms have turned out to do exactly the same class!
of things — to have the same general calculating!
power. This suggests the very powerful thesis oil
Church that what these mechanisms can calculate!
is all that can be calculated, no more, no less. Con-ji
sequently, a study of this class of devices — device.')

which can calculate the so-called recursive func-ll

■

tions — might, in fact, be a proper study for one whcjj
wants to know what machines can do and what the)] |
cannot do in principle.

Another important kind of investigation in thtj
field of symbolic logic which dates quite a bit farthei J
back, is the attempt to formalize within formal Ian*
guages the inference process that we perform irj
natural languages. People have attempted over cl
period, at least as far back as the 1800's, to reprej
sent within the language of formal logic the asser
tions and the inferences made by a speaker of i\

natural language. Interestingly enough, there is
some reason to believe that this quite unsolved prob-
lem has remained unsolved because the complexity
of the problem is' such that a solution could only
be implemented with something of the complexity
of contemporary machines, which suggests why the
logicians of past decades, although they had con-
siderable insight into the problem, might not have
been able to create the solution that they were look-
ing for because in this case the necessary mechanical
devices, electronic computers, were not available to
them.

There are questions being asked in the field of
Automata Theory. For example, how one can de-
sign them efficiently in a sense that is independent
of the particular device used for constructing them,
and what kinds of strings of symbols specialized
machines can produce or analyze. A rather exten-
sive literature in this subject of Automata Theory
has developed. An interesting recent development
in this area is the possibility of machines which can
in some sense organize themselves to perform very,
very elementary tasks. This is a highly controver-
sial matter at present.

This leads me to the last and by far the most
controversial subject, except to those engaged in
such research — the subject of Artificial Intelligence.
The earliest question asked in this area was asked
by the British logician, Turing, when he asked
whether machines could think. I certainly will not
attempt to offer any opinions on that subject because
one immediately becomes involved in controversies
over what is meant by thinking, what is meant by
intelligence. One finds interestingly enough that
as each little problem in this area becomes solved,
it becomes generally agreed that that is not of
course, intelligence. So, the goal of intelligence
manages to continue escaping us, but just by defini-
tion. I might mention to you some of the things
that are being done in the area of artificial intelli-
gence. There are pattern recognition devices which
can read printed characters, and there are investi-
gations concerned with very much more complex
kinds of information for example, electrical circuit
diagrams.

There are investigations in the artificial intel-
ligence area concerned with problem solving, the
Intention here being to attempt to model in a func-
:ional way the kinds of processes carried out by
human beings when they perform such tasks as
Droving theorems in elementary plane geometry,
aroving theorems in logic or solving very simple

creative tasks. It has been shown quite con-
vincingly that one can program a machine to ex-
hibit a problem-solving behavior, at least superior
to that of the person who programmed the machine.
For example, one investigator has programmed a
machine to play checkers against himself and to beat
him. Others have programmed a machine to prove
theorems in plane geometry, and, in one case to
exhibit a very ingenious new and creative solution
to a single geometry problem, one which I believe
no high school student had previously come up with,
even though it is self-evident as a beautiful elegant
proof, once you see it.

In the area of artificial intelligence, there has been
a great deal of interest on the part of many people
for some time, in the subject of game-playing
machines, such as the checker-playing machines.
There is a great deal of active interest in the possi-
bility of programming a computer to play chess.
There is no chess playing program in existence today
which compares in ability with the checker-playing
program that I mentioned. And that is simply a
reflection of the fact that chess is a more compli-
cated game.

Finally, still in the area of artificial intelligence,
I might mention investigations that people have
been conducting into the subject of inductive in-
ference — how we might get a machine to deduce
consequences that do not strictly follow with re-
spect to any ordinary deductive systems — conse-
quences that follow, but in an inductive way, from
certain premises. For those of you who are in-
terested in the fine delicacies of this question, it is
evident that once you write a program for carrying
out inductive inference, then you have formalized
this process, and you are doing nothing more than
carrying out deductive inferences. You might say
that the main goal of people in the artificial intelli-
gence area is to take all these problems that seem
unformalizable and to trivialize them by making
them formalized and by attempting to show that
they can, in fact, be solved very simply. The
ubiquitous phenomenon is that once we have solved
a problem, it certainly appears simple.

I could elaborate greatly on the technical details
of some of these subjects but instead I would re-
emphasize the fact that all of these investigations
are being carried out generally by highly competent
scientific researchers. The results of this research
are generally communicated very effectively
through the usual vehicle of scientific publication.
They are, of course, available to the international

201

community. The results arise in every nation of
the world and, at least with respect to this aspect
of research, the mechanism exists today in a highly
developed form, for communication and coopera-
tion.

Again, I should mention that for those of you who
are concerned with the possible automatization of
patent searching, with patent analysis, with the
patent issuing process, the option is available to you
to become sufficiently familiar with some of this
highly theoretical research to be able to decide
whether or not to engage in this research or perhaps
to follow it closely, so that the option will thus be-
come available to you to help influence its direction
in such a way that its eventual application will be
in the area of the patent problems, rather than in
other areas chosen by the people conducting this re-
search. This option is available to you and is a
costly one. It offers the very enticing prospect that
we will be able to do what any good scientific in-
vestigation enables one to do, namely, to accumu-
late experiences that the second investigation may
build upon the results of the first investigation, so
that the consequences become more and more pro-
found as one progresses in this area of scientific
research. Thank you.

Mr. Glaser: When I said that we proceed at a
merciless pace, I did not mean it was merciless on the
speakers. They are used to it. It is merciless on the
audience. There was an indecent amount of mate-
rial covered in this talk, of course. I hope that when
the proceedings come out, which is not too far away,
those of you to whom most of this material is new,
will have a chance to look it over and study it a little
more in a little more detail. What Mr. Kirsch was
speaking of, of course, is the under-current of grow-
ing technique and theory which we hope will de-
fine the eventual achievement possible in these kinds
of problems. We will, however, try to rush the pro-
ceedings into your hands so you can look at these
things in a little more detail and see in what con-
text for example that thing he said about proving
Euclidean theorems came up.

We now have another, I am afraid, unduly brief
presentation by Larry Heilprin. The title given this
presentation is "On International Collaboration for
a Complete Information Storage and Retrieval Sys-
tem." Dr. Heilprin has his doctorate from Harvard
in physics under the well-known physicist, Professor
Bridgeman. He was at the Bureau of Standards for
ten years, so you see we people from the Bureau seem

to sneak into the program in various ways. He is
now at the Council for Library Resources, Incorp-
orated, which is, in itself, such a complicated story
that all I will tell you about it is they somehow seem
to be able to get some money to give to people to
do research. This is a very useful function. Dr.
Heilprin has about 20 minutes, I think, to make his
formal presentation. We will ask everyone to hold
questions until the end, so that those who must leave
at 5 : 00 o'clock will at least hear the three presenta-
tions. Larry, the meeting is yours.

Dr. Heilprin: Ladies and gentlemen, what shreds
are left of new things to say I will try to compress.
Almost everything that I had to say has been partly
said, but I would like to, shall we say, take off at the
point where the preceding speaker left. Only, my
interest is somewhat different from his. I am inter-
ested in the basic philosophical background of the
whole picture. And I would like to place big can-
vasses, and of course sometimes I get mixed up in
doing this.

The first thing I will do is to indicate what I think
is a complete information retrieval system. If we
take the two types of communications which exist —
the so called short duration communications which
last up to a few minutes, and which employ time-
modulation principally, and the long-duration com-
munications which, in order to endure anywheres
from a few minutes to many years, employ solid
media and spatial modulation, we immediately have
a spectrum in which the path of propagation of all!
communications is broken up into two parts — a short
part where the time axis is quite small for the s.d.i
or short duration, and all the rest of space and time
for the l.d. or long duration communication. Nowfi
that is the one with which the documentalist is con-j
cerned — l.d., long-duration communication. That J
is my first picture.

Secondly, the fact that this time axis extends,!

. , il
indefinitely means, of course, as we have all realized.!;

that the mind that puts the initial modulation dowrjl

may not necessarily be contemporaneous with the,

mind that decodes it at the end. And this is the!

principal problem that makes long duration com- J

munication a very specialized kind of thing as com-i

pared with what we ordinarily call communication j

engineering. It is far more complex. It has alj

kinds of different, shall we say, occupations, from

writing to publishing, to operating libraries, being aj

documentalist, running specialized retrieval systems

such as patent offices, and so on.

202

As a background for this paper then, we have this
long duration communication and we have mind
to mind communication in which there are a set of
main stages or states through which the l.d. message
passes. And I want to point out that what we call
a library or a patent office is simply a sub-set of these
stages, of the paths of propagation in four dimen-
sional space, that the inputs are not controllable in
the sense that the internal processes are. Well now,
having talked somewhat vaguely like this, the title of
this paper could apply to this larger system. It could
also apply to patents only, and what I want to do
is to fit the patent picture into the larger picture
to show how tremendous the big problem is and how
somewhat more simplified is the smaller picture for
the patents. And finally I want to bring it around
to the point of international collaboration.

We have talked about a complete system, which
means from one mind to another, in my definition.
Now, in another paper which has not yet appeared,
the American Documentation Institute is getting out
its version of what documentation is, and this is
the general picture. The documentalist is a com-
munications engineer-long duration communica-
tions engineer — whose function it is to get these
messages from the sender to the recipient and he
does so using any means whatsoever that he can.
But in general, these means break up into two chan-
nels — purely physical, or shall we say, mainly phys-
ical means, such as carting books around and put-
ting them on shelves and marking them and binding
them, and mainly intellectual tasks, such as clas-
sification and reference work, and searching ideas.

Now if you break up a library into a flow diagram
and do not insist upon a single flow diagram then it
would look something like this. I would say here is
your mind No. 1 and here is your mind No. 2 down
there. We have some intermediate stages, and then
we have a library which is equivalent to, we will
say a patent office system which is a specialized li-
brary, and we will have two lines of flow. I'll make
it terribly simple, by calling one the intellectual proc-
esses, and the other the physical processes. The
latter will be to collect and to store and to dissemi-
nate, and their counterparts in the intellectual proc-
cesses are acquisition, search, and reference. Now
they are not all straight though. In fact, there is
a loop in a library, between distribution and store,
as you know when you return a book. And I am
merely suggesting that a library has a set of tasks
which are primarily physical and it has another set
which is mainly mental. Now the physical tasks

center around the enormous growth of libraries or
patent offices, or shall we say in general, these l.d.
messages. It is very common to use the expression
"they are growing at an exponential rate". How-
ever, as any of you know, if the exponent is small
enough, the curve can be flat or it could go down.
So the use of the word exponential means nothing.
What is meant by exponential growth, of course,
is that the exponent is large. And we might say that
these libraries are growing something like this, *N =
Nn.e" 1 where (S is some respectable quantity, and the
time in order to double is on the order of anywheres
from 5 to 25 years. Some big institutions, such as
Harvard, have about 1 7 years time to double.

Now here we have two channels for the library,
and the question is, for the immediate future, not
for the remote future, where in those sets of very
big and rough classes of operations would you be
able to, shall we say, apply the methods of systems
analysis, operations research, automation, and real
payoff in mechanization. Where could you in that
system? As Mr. Kirsch has pointed out to you,
there is in this flow line practically no existing hard-
ware. There are wonderful experiments that are
being performed, but nobody has done anything
yet that actually pays off in the sense of large-scale
operations. Therefore, I would say, if you want to
make a broad generalization, that the first line of
flow for all the intellectual processes cannot at the
present time be automated. This is a very sweep-
ing statement, and I know that there are many
people who would like to take exception to that.
Nevertheless, I am saying it, with the saving phrase,
"for the immediate future". Therefore, I believe
that automation of a large-scale system, such as a
patent system, and eventual world collaboration, for
that matter, if we are talking about automation,
will have to stick to these physical processes. Now
these physical processes — the first one, "collection",
is linked to the preceding stages that come from
this line, maybe a publisher or distributor, and so
forth. It would be pretty hard to automate the
process of "collection" because you would have to
change the way publishers act and the way the mail-
man delivers, and so forth and so on. So, my
picture of where automation is going to take place
and can take place is in the mechanical or physical
aspects of storage and dissemination. This is where,
also, the greatest hardware advance has been made
that is either actually now here or very close to
completion. As you all know, there are a number
of systems that are coming into being. The first

203

one of course is here — the Minicard system. Mag-
navox has a fine system, and IBM, of course, has
announced its Walnut system. The AVCO Cor-
poration has developed a store and dissemination,
but without the brains that go into it — that is to
say, you can use anything from a card file to a com-
puter to find what is in it. However, they have
combined one-second retrieval with great scale re-
duction. They can get a million pages into about a
cubic foot or more. In general, there are many
other devices that I am sure are as good or better.
I am simply saying that in this area there is hard-
ware for storage. Secondly there is hardware that
is available for copying and rapid dissemination,
which will solve the problems, for instance, of Mr.
Hurd of this Office. How is he going to dissemi-
nate 150,000 patents per day or something like that
of which only 10 percent goes to the outside and
90 percent goes to the examiners themselves.

Dissemination methods, then, can be automatized.
So, now we have a picture of, shall we say, a single
one of these units, and the most likely area. Cer-
tainly, I am not saying that something may not be
found here by Mr. Kirsch or others that will com-
pletely revolutionize this — some automatic indexer
or something. But, right now, I would say, those
two areas which I have indicated are the places
where some real progress will be made and where
it is about to be made. Now, what will happen?
Supposing that we take as an example the system
that I know about the best — the AVCO system,
where photo storage is used. Instead of physically
translating a volume out of the library and bringing
it back into the store, we will eliminate the loop
between distribution and store and it will be a
straight through system, in which the price of the
copy is below a certain economic threshhold —
namely, it is dispensable. This is an idea that is
very common. Nevertheless, it is true that the cost
of the l.d. message has gone down in history since
pre-historic times. If I may draw in the air, if this
is the time axis, and this is the present, and this is
the cost in man hours of enscribing the number of
bits in a modern book, and we use some sort of a
log scale, the price would be pretty high if you have
to enscribe that book by chiseling on a rock, or even
on bone or even if you got it down to paper and
did it on paper or papyrus or a scroll. The human
labor of scrivening and writing would be so high
that over thousands of years this cost would fall
very slowly. Then it fell sharply when we had
movable type. But the cost of that book in man

hours has now crossed an economic break point in
two directions. The first one, of course, we are
all aware of. In order to see what that point is,
let me go back. The cost of circulating this loop
is now calculated by librarians to range from $.25
per circulation to $1 .00, something like that. There
is a tremendous amount of cost in charging the
thing out and charging it in, keeping records of
who has it, finding out what is happening, who lost
it, how much they owe, computing it, chasing after
them, replacing the book when it is lost, and so
forth. There is a corresponding loss by the user,
which is, of course, loss of time and accessability.
At any rate, this loop will be eliminated and the
user will have a "D" Library, as I call it — a dupli-
cating library, as against a circulating library.
And, in fact, the library merges with the publisher,
because the publisher who now stocks a huge inven-
tory can do the same as the library. He can have
a single master copy as Mr. Hurd, I hope, will
some day have, and he will duplicate beautiful
prints of patent number 2976 and so on, which he
can retrieve because it is a beautiful serial number,
and all he does is punch that on a paper tape and
out come the six sheets. On demand publication.

The publisher may say, in self-defense, "I will also
be a librarian. I will also stock my books." And
he will also issue them on demand, and he will no
longer have to have the cost of inventory of tre-
mendous numbers of books that have to be kept in
a warehouse. So, I would say that the future, then,
of the library and partly of our patent system, is a
compression in which we will have publication on
demand, and this is all in the physical realm.

Now the cost of the intellectual work, which I be-
lieve is the thought that Dr. deHaan had, the actual
sharing of the classification work so as to bring it
about and make it uniform throughout the world, is
very much greater than this. A breakthrough prob-
ably is not yet in sight, and while I personally feel
that I would like very much to participate in push-
ing this effort, I would say that for the immediate
future, its hope of getting standardization enough to \
really help, is somewhat more remote. But possibly
computer methods and programming, and so on, i\
may do it.

I had a lot more to say. I won't say it.

Mr. Glaser: The final speaker in this session is going
to address herself to a problem that has come up
very frequently in the conference. It is that of auto-
matic reading and picture recognition. I could tell I
you a great deal about Mary Stevens, but the thing

204

you want to know now is that she has just finished a
report on this subject for the National Science Foun-
dation. We will see to it that one of the first acts of
the new, informal committee on international co-
operation will be to mail each member of the ad hoc
committee a copy of this report, which is just avail-
able now. Mary Stevens will now summarize the
present state of the art in mechanical reading of
characters and of pattern recognition. Mary
Stevens of the National Bureau of Standards.

Miss Stevens: Well, the report contains something
in the order of 75,000 words, and to recapitulate this
would keep you here far longer than your planes,
trains, and so on, would permit. Therefore, I shall
try some short cuts and will try to cover as much as
I can to give you some idea of what we think the
current state of the art in automatic character recog-
nition, in fact, is. I think I could summarize best by
saying that the current state of the art in automatic
pattern and character recognition is in a "4-F sta-
tus." For the benefit of our internationl visitors, I
might explain that 4-F is one of those classification
terms that got assigned to draftees who were not
accepted because they were considered unfit for
military service. Well, our 4-F status as far as char-
acter recognition is concerned, does not mean this.
I will describe these "4-F'S as being Fiction, Fallacy,
Fact and Feasibility.

Let me start off first by saying that it is a fiction,
I think, to consider that automatic character recog-
nition is not ready for service. And it is a fact that
in limited circumstances, under controlled condi-
tions at least, character recognition is ready for at
least some limited service. Now the other aspects of
the fiction F, like the output of popular fiction, with
which information retrievalers are not as yet con-
cerned and perhaps hopefully never will be, ranges
the gamut. The pessimistic school, such as that of
the Southern sharecroppers living in misery, is rep-
resented by those who say, "Oh well, manana — five
years from now," and they sit back, shall we say, in
the squalor of the present rather clumsy methods.
Then there is the existential school. You know, this
is the Doubting Thomas: "Until I see it for myself
and until it happens to me, it does not exist." Then
there is another school which is just as mistaken in
its own way, and that is the optimistic one — the
romantic kind. This is the school that says the
punched card is already obsolete. Well, whether the
punched card ever will be obsolete is a moot ques-
tion. There are, however, as I hope we will see,

situations where certainly there are other techniques
that may be preferable and may indeed be economic.

The fallacies that mark current progress in the
field include an apparent fallacy on the part of
many potential customers who seems to feel that it
is necessary to solve the problems of reading cursive
handwriting before one can even consider applying
machines to the reading of type and printed text.
This is a fallacy. It is a fallacy because, for one
thing it, overlooks the fact that certain elaborate
costly computer programs can already succesfully
read human hand-printed material almost an well
as human beings can. In other words, the cases
where the machine is confused are almost precisely
the cases in which human beings seeing the character
out of context would be equally confused.

A second fallacy is what we might call the super-
man fallacy — the potential customer says, "This is
all very well but I cannot afford error rates or reject
rates such as you tell me would be necessary; I
cannot afford more than one error in a million
characters." Now, what does he live with today?
Even in banking operations, he lives with two errors
in 10,000 after verification of punched cards, ac-
cording to an independent survey made by Stanford
Research Institute, when it was embarked on the
ERMA program. When we are asking for a
machine to go 100 to 200 to 300 times faster than
human fingers, at a cost that is commensurate to
1/1 00th to 1 /200th of the cost, why should we also
ask the machine to be more accurate by a factor of
10 or 100? In other words, the fallacy here is to
expect this superhuman accuracy rate when, in
fact, we can live with no worse than we get today
and still make money and gain time, and also pro-
vide a consistent kind of output that is easy to check.

The third kind of fallacy has at least a grain of
truth in it. It is that optical character recognition
may be considered the Cinderella of data process-
ing. The fallacy arises, however, in assuming that
this Cinderella means Cinderella in the sense of
waiting for a fairy godmother. Rather, the Cin-
derella that optical character recognition represents
today, is Cinderella waiting for Prince Charming to
come and bring the slipper which sure enough, her
dainty foot will fit. So, to turn to fact then, I
think that it can best be illustrated by the ceremony
held under the auspices of the U.S. Patent Office
several weeks ago, in which the 3,000,000th U.S.
patent was issued. This patent was for a reading
device ! The choice was deliberate ; the applica-
tions likely to be processed at the time were care-

205

fully considered, and that particular one was chosen
which was believed to represent the greatest impact
on the American public, on industry and on govern-
ment. It was for a device to read magnetic ink —
and most American citizens by now are thoroughly
familiar with this on their personal and other checks.

Actually, the magnetic ink field is the one in
which there has been the greatest practical use made
of automatic character recognition. Technically
speaking, however, within the same limitations of
highly standardized and not very elegant characters
and of just the ten numerics and four special sym-
bols, it is just as easy, just as economic, and so on,
to do this recognition optically as by magnetic tech-
niques. The reasons for the choice of magnetic
ink relates to other characteristics of checks, such as
the fact that people spill coffee and stick their chew-
ing gum on them. You need to read through con-
siderable obliteration.

Another fact in optical character recognition is
that there have been a number of deliveries of work-
ing devices. There are some four of five companies
here in the United States that have delivered a pro-
duction model, and at least one abroad — Solatron.
The pioneer in this field with respect to deliveries,
Farrington Electronics, formerly IMR (Intelligent
Machines Research ) has over 50 machines of various
types in customer use, and these range all the way
from specialized numeric font machines to a page
reader. There is a teletype reader delivered by Far-
rington that is being kept so busy in a communica-
tion center that the research staff who ordered it have
hardly had time to look at their "baby."

More significantly, since the completion date of
May 1961 for the report which covers, or at least
mentions, some 160 different equipments, simula-
tions, ideas, patents, and so on, many developments
have occurred. In the last month or two for ex-
ample, a colleague and I have visited some of the
possible suppliers, and we are reasonably confident
that within the next few months we will receive
favorable responses to a bid request, not for re-
search and development machines, but for service
bureau type of contract. We would supply them a
facsimile copy, photo or xerographic or otherwise,
of a Russian journal, and they would deliver to us
magnetic tape. The cost would be on a strictly serv-
ice bureau basis — so much per word, so much per
line, so much per page. We expect at least six
companies could respond to such a bid request.

Where you can define your specific requirements,
where you are willing to settle for approximately

the accuracy that you get with keypunch opera-
tors today, where you can assure that the quality of
the input material will be good and will be quite
consistent, the fact is that optical character recogni-
tion is with us now.

Now, feasibility, the fourth F, means that there is
available equipment for specific special requirements
and for realistic performance specifications. That
does not mean, as yet, that there is a universal reader
which can read anything it has not seen before. It
is quite possible that this will never be the case. But
a second aspect of feasibility is that there is very
promising research indeed, not only in terms of more
generalized machines, but also in terms of the broad
area of pattern recognition, which will extend to
graphic shapes, to diagrams and potentially, we
hope, to at least simplified pictures.

Now, briefly, how is automatic character recogni-
tion acomplished? There are a number of tech-
niques described in more detail in the survey report
and in many other accounts in the literature. The
oldest technique in point of time of proposal seems
to be the idea of a photographic negative image
which is transparent — clear where the A would be
and opaque elsewhere, and similar masks for the
other characters of the alphabet. When the charac-
ter image is presented to the machine, it is reflected
in such a way that the black on the paper is super-
imposed on the transparent part of each mask. In
exactly the one case where the character is found
to coincide, namely, when the A mask is being looked
at, there will be extinction of light to a photocell
behind it. We thus recognize A. If we show this
A mask to an input character B, there will be white
coming through in some areas. When we couple
this kind of negative mask with a reversal of the
image or positive mask, we can determine many
things. Note that for such characters as Q and O, if
the character is not well-formed, the tail of the Q
may get lost. For this reason then even in this so-
called photographic template technique, many
modern versions use a form of weighting. They
consider the blackness or the whiteness in certain
spots much more critically than in others in order to
arrive at an improved separation criterion.

In another current method, the image field can
be split up by scanning techniques which in effect
divide the field into a number of discrete cells.
And we can require that for any cell, if 50 percent
or more of the area scanned in that cell space is
black, we will consider it all black. Now again
we can, in effect, use a template for recognition.

206

We can simply say that for these cells considered
as elements of a matrix, cell one, the first row, the
first column, must be black or must be white for a
given character. We can do this for all the cells
that might be covered, or we can do it only for
selected cells that are systematically chosen to dis-
criminate between the characters of our alphabet.

We may do other things. We may divide the
character up into zones or regions. Consider, for
example, the field in which the character is. We
simply scan this field vertically and convert any
black we see in any one vertical scan into an equiv-
alent wave form. If this particular vertical zone
were seen as all white, we would have no "black"
output. When we get to a vertical zone in which
some black appears, we begin to pick up a little.
Now in a succeeding scan we get quite a lot of
black, and we would have a characteristic wave
form that could be used to identify the characters
of a small vocabulary. This technique, by the way,
the characteristic wave form method, is that which
is commonly used in most of the magnetic ink char-
acter recognition techniques.

We may also do other things. There is not time
to talk about them all, but let me mention one that
gets away from the fact that you have to have pre-
viously seen examples of the characters that you are
going to recognize in these so-called template tech-
niques. If you have a very thin input character
A and your mask is for a broad A, obviously it
will not recognize that A. If the A is an italic A,
it will not fit either the photographic type of mask
or the A which specifies the cells to be black and the
cells to be white for the character A. However, if
we do scan successive vertical zones and determine
for a given segment that there was a continuous
long black leg and that at some later point there
were three short black bursts, we arrive, in effect,
at what might be called a machine idea of an E.
The example given here is called stroke analysis, but
it is also true that similar techniques can be worked
out in terms of enclosed areas "lakes," partially
enclosed areas or "inlets", and other characteristics
that tend to be invariant with certain types of
characters, all of which we can call "A" or "E"
in some alphabet or another.

This, of course, does not as yet help us with hand-
writing. I think the point to note, however, is
that in many cases these techniques and others
such as curve tracing or vector crossings, are not as
important in determining practical feasibilities today
as is the exact size of the vocabularly — the total

number of characters you want to recognize and
the consistency with which you can guarantee that
they will be presented to the machine in "good
shape" — or, contrariwise, the extent to which you
are willing to take reject rates of 20, 30 to 40 per-
cent. By the way, you can still make money. The
reason for it is the insistence on good quality or a
high reject rate.

When we look at a blowup of typewritten charac-
ters we see holes, smudges, filling in of closed space,
and breaks in character strokes. When it comes to
carbons, it is much worse. They are not clean, good
characters. Nevertheless, typed material can be
read if you will accept the higher reject rates. Again
consider the human factor. In some tests that were
made for the U.S. Navy on typewritten material pre-
pared in the field by a variety of shipboard person-
nel, there was a high reject rate. It was around 33
percent. That sounds unacceptable. Actually, they
could still make money with that reject rate. What
is more important, however, is the fact that over
25 percent of the total input was tossed out on a
manual sort. Why? The instructions to the typists
said — "Do not make strikeovers." I wish you could
have seen what they actually did. So a high reject
rate is often the result of the simple failure of per-
sonnel to follow instructions. Sometimes they
xxxed out the same character as many as three
times.

We have thus taken a few quick looks at the
"hows," of character recognition. There are many
other techniques. There are intriguing methods pro-
posed that go into the use of random operators of
various types. Some of you may have heard or will
be interested in hearing further about the Bledsoe-
Brown ring technique in which cells in a matrix are
taken at random. For a given number of samples,
this pair or this triplet of randomly chosen cells is
or is not black, and the procedure is repeated for
other characters in the vocabulary. This can be ap-
plied to handdrawn characters. At MIT, work that
is is along the criterial analysis approach — which I
have illustrated in the case of stroke analysis and
"lakes" and "inlets" has again been used on hand-
written characters and this with a score in the neigh-
borhood of 87 percent.

Curve tracing and topological techniques have
been explored by Sherman and others here in the
United States and by several of the Russians. There
are many other "hows." The point is merely to in-
dicate that there is a wide range of ways to recognize
fixed, formalized and relatively standardized pat-

207

terns. These available techniques will not as yet
help us recognize a configuration on an aerial
photograph, where the exact geometrical or topo-
logical pattern of runways to identify an airport, may
not be known in advance, where we will have to
evolve rules for helping us to decide that over a period
of time. For example, if there are certain straight
lines joining at certain angles and certain configura-
tions, maybe somebody should take a look at the
original because maybe an airstrip is shown.

Possibilities for pictorial data recognition suggests
some of the whys for automatic character recogni-
tion and the "why" of talking about it in an informa-
tion retrieval conference. It is to be recognized that
character recognition is an aid to input. As such, it
avoids the question that was bruited about interest-
ingly yesterday, whether we should have full text
processed by machine or whether we should have
human abstracting, because these techniques of
character recognition can be applied to the full
text if that is what you want and also to the prod-
uct of the analyst's mental effort — what he writes
down. And since various types of recognition tech-
niques can be quite readily extended to graphic and
geometric shapes, it may be that what the analyst
writes down will be a notation that is more familiar
to him than other codes that he might use. For
example, he might use shapes or symbols familiar
either in terms of circuit elements or chemical struc-
ture elements. This would be read immediately by
a machine. It is also quite feasible, I think, to
look ahead to automatic reading of precisely the
kind of flow chart that Mr. Wolk talked about
yesterday — that is, if he would be good enough to
perhaps help us a little by making certain kinds
of boxes squares instead of rectangles, so that we
would tell them apart in the reading.

As a third possibility, beyond the text input and
the encoding of the analyst's notes, however, pattern
recognition provides a ready means for further ex-
perimentation and research. This means both for
statistical analysis of linguistic phenomena in text
and for the kind of exploration of picture processing
technology that Mr. Kirsch touched on briefly.

In conclusion then, I think we can say that optical
scanning techniques do offer and do promise im-
mediate usefulness in some areas as aids. The re-
search aspects point to what are really common
interests in pattern recognition. The elements that
make you want to choose this document and not
this comprise patterns whether those elements are
words, names, or the ideas expressed by them, or

the way they come together' in concatenation, or
again the flow chart approach. Mr. Wolk, I think,
points to the same idea, that there is a problem in
pattern recognition whenever we want to search
and to select, and these two areas of recognition and
of selection can feed one another.

Finally, I think we need to realize, when we think
of 7,000,000 patents, that human efforts themselves,
either on the mechanical or the intellectual side,
are not good enough if we want to move forward
in mastering this exponentially growing stockpile.
Therefore, we are going to have to increase the
degree of collaboration possible between the man
and the machine. Automatic pattern recognition
techniques can help us to do this. The discussion on
terminology this morning reminded me of the fact
that at the National Bureau of Standards all sorts of
inquiries are received. One day I picked up the
phone and a gentlemen introduced himself, so and
so from such and such a company, and he said, "I
understand you people at the Bureau can tell me all
about information "retrieval." I raised a mental
eyebrow and we talked briefly. Actually, I think
perhaps the term is appropriate. We are going to
need to try — many are going to need to try — and
this means the researcher and the linguist and logi-
cian, as well as the examiner and the attorney and
the documentalist and the engineer. We are all
going to have to try and we may have to keep trying !
and trying again. Thus all that can help us should
be looked at seriously without forgetting the other
point that has been made so strongly — the fact that
there is a major part of the iceberg still under the
water. The part under the water is, of course, the
question of deriving useful meanings from messages,
pictures, communications, and information of all
sorts.

Mr. Glaser: Thank you, Mary. It would really be
quite a problem, I think, to engage in any systematic
discussion of the enormous amount of material
covered by the three speakers in this last session.
As is our custom, we will excuse those who have to
leave at 5 : 00 o'clock because of fixed transporta-
tion arrangements. If there are questions on any
of these papers, this is the time to entertain them.

Mr. Jeffery: I would like to set a problem to our
character recognition expert in the hope that when
we return within another few years, we might have
the answer. That is, can we have a machine for
searching picture trademarks? That is quite a
problem in trademark offices. All we want to do is

208

to pick out pictures which bear more or less re-
semblance to the picture in question.

Miss Stevens: I think this goes to a question that
we really had very little time to get into at all. That
is the question of the research into the relatively
invariant features of symbols or characters, that we,
as humans, would call the same. It is the problem
of recognizing the script A, the upper case A, the
lower case A, the italic A, the gothic A, and so on,
as all being A. By the same token, for your trade-
mark case, you would want to recognize a bell as a
bell, regardless of its size, its proportions, it position,
perhaps in terms of other material and similarly,
for all sorts of other things. This is in an area in
which I would be perfectly happy if 50 percent of
my searches could be done this way, and the rest
would have to be done other ways. It is quite
probable that present-day research could hopefully
help you out on some 50 percent of the searching,
in say two or three years, providing you are willing
to pay the bill for research and development.

Mr. Glaser: We have now had a number of people
who have had to leave. At the risk of seeming
arbitrary, I would like to close the meeting formally,
the rules always being that if you can catch a man
before he gets out you can ask him a question. I
think we have exhausted the members of this con-
ference. They have been very patient to sit
through a very long and demanding group of ses-
sions. I think everybody's patience is much to be
commended. This has not been easy on them. I
would simply like to close the meeting now with a
reminder that our foreign guests are to be welcomed
at the Patent Office tomorrow morning, where the
Commissioner has a program planned. I believe
that all of you, both our foreign visitors and people
in this country, will probably hear a great deal more
of this ad hoc committee, as it goes into operation
and starts to perform what I hope will be very
useful services. Thank you.

The last session was adjourned at 5:00 P.M.

209

APPENDIX A

PLANNING A LONG-RANGE PROGRAM FOR MECHANIZED

PATENT SEARCHING

Martin Kalikow
General Electric Company

Presented Before the Patent Lawyers' Club, Washington, D.C., October 24, 1961

Ladies and Gentlemen :

The patent system as we know it today is engaged
in a life and death struggle — not so much from its
external enemies, but more critically from its own
internal information weight and faulty information
digestive system. It has already suffered several
minor heart attacks and will soon receive a mortal
coronary if we do not find some way to improve its
capacity for information absorption, circulation and
selective retrieval.

The crisis is fast approaching ! If it continues to
take several years to obtain a patent; if the Patent
Office becomes unable to find the best prior art in a
large percentage of the litigated cases; — if companies
begin to find it cheaper independently to re-invent
the wheel than to conduct an expensive and un-
reliable patent and literature search ; — in short, if the
patent system, as a result of inadequate searching,
should belatedly spew forth invalid property rights
and should no longer be able effectively to function
as a repository of retrievable scientific and technical
achievement, it will deserve to be sent to its demise !
The patent system is supposed to promote the ad-
vance of Science and the Useful Arts, not confound
and stifle such advance by an impenetrable and
engulfing volume of words !

But let us not despair. Words have gotten us
into this mess, and words can get us out. By this I
mean that the various ephemeral "ideas" of a patent
that we may be seeking are locked within its words
(and also within its drawings — but principally
within its words ) . What must, therefore, be ex-
tracted, collated, compared, and otherwise manip-
ulated by our machines is the "meaningfulness"
embodied in these various combinations of 26
letters. And before we get into the meat of our
discussions this evening, it is important that you
have some understanding of the semantic structure
of our English language. What our experience in

the use of a "basic language" type code has taught
us is that, just as the 26 letters of the English alpha-
bet arbitrarily represent certain relatively basic and
phonetically distinguishable sounds of the human
voice, in this same way there are certain words in
the English language which represent certain rela-
tively basic and semantically distinguishable con-
cepts of the human mind ! — and this is true regard-
less of the specific language used, — whether it be
English, French, German, Russian, or some other
language.

In other words, just as there is a continuous
spectrum of sound gradations which have been arbi-
trarily carved up by common agreement in accord
with certain alphabetical identifications, so in this
same way there is a continuous spectrum of mean-
ingful concept gradations which can be arbitrarily
carved up by common agreement in accord with
certain word identifications. Complex words,
which convey complex meanings may then be fac-
tored with some semblance of uniformity into their
component more rudimentary (even though not
ultimately basic ) meanings. For example, the word
"spectrophotometer" may be translated during both
classification and retrieval steps into its component
more rudimentary meanings represented by the
words, "light, frequency, select and measure".
However, the trouble lies not in the actual perform-
ance of any one of these translations from complex
to simple words — each can be done by any intelli-
gent mind quite readily. The great difficulty lies
in the problems of consistency. The fact that no
two human minds will make precisely the same
translations of the same complex words, and that
even the same human mind will not make precisely
the same translations at different points in time.

Fortunately, it is just in this area of "consistency""
that the more sophisticated machines can save us.
Once a given complex word is programmed to be

211

translated in accord with a given semantic code into
the storage of large computer, the computer will
always make exactly the same translation whenever
that same complex word is encountered thereafter.
If experience should indicate that the translation
was not accurate or not exhaustive of the meaning-
fulness of the original complex word, it becomes a
relatively simple matter to introduce or substitute
other basic words in the machine translation
process.

Keeping this analysis of the function of machine
processed language in mind, I would like now to
back off a moment and survey the problems in-
volved in planning a long-range program for mecha-
nized patent searching.

First of all, let us not be too pessimistic. The
tremendous volume and intensity of investigatory
and developmental activity during the past few
years in the field of information data processing has
provided the bedrock upon which a long-range pro-
gram of mechanized legal and patent searching may
now be built with some assurance that the edifice
will not come tumbling down. Sufficient experi-
ence with specialized information storage and re-
trieval systems has been acquired that certain
common denominators for a generalized system are
emerging. It is now becoming possible to evaluate
systems in accord with their contributions to the
evolution of the art (or science) of documentation
and to select those features of each type of system
which will probably need to be incorporated in any
large future system. Moreover, the corresponding
evolutionary trends in computer technology have
become fairly well defined so that we can extrapo-
late toward expectant "computers of the future"
and tailor any long-range program to take advan-
tage of and be compatible with future equipment.

The time for making the requisite analysis and
for planning the main outlines of large searching
centers has also arrived. Most professions whether
technical, legal or medical have now been aroused
to the acute problems inherent in the engulfing vol-
ume of published material, and the imagination of
the public as to the potentials of mechanized infor-
mation data processing is being captured. Money
in fairly large quantities is being proferred from both
public and private coffers and will be forthcoming
in even larger sustained quantities as programs are
tendered having good promise of long-range success.

Any such long-range analysis at this stage of the
game is, of course, fraught with many dangers of
omission, overemphasis, unpredictable technical or

linguistic break throughs, etc. ad infinitum. How-
ever, despite these risks, it is believed worthwhile
to make the attempt particularly if by doing so,
others are led to challenge some or all aspects of
the programs suggested and thereby give impetus
to a sounder professional consensus as to the proper
directions or emphasis. Accordingly, I shall hereby
stick my neck out and set forth for your evaluation
what I believe to be the technical requirements
which must be met in any long-range program for
mechanized legal searching.

At the outset, let me state what I believe to be
some discernible trends in information processing
computer equipment. Based on the recent history
of computer development, it is believed fairly safe
to predict continuing developmental achievement
in the near future toward:

1 . greater concentration of bit density on the record-

ing medium

2. greater storage capacity in the computer itself

3. faster random access to stored data

4. higher speeds of data manipulation and compari-

son

5. higher speeds and reliability of input into and

output from the recording medium

6. machine conversion (for example, by optical

scanners) of ordinary printed text into ma-
chine usable recorded form

Also predictable of achievement but at a some-
what more remote future date is direct machine
conversion of all types of "images" such as photo-
graphs and diagrams into machine usable recorded
form as well as direct manipulation and read-out by
the computer of the image itself.

Now, let me set forth what I believe to be some
of the basic "documentational" and "operational"
requirements for large searching centers.

Firstly, there must be "comprehensiveness" in the
creation of the initial "master recorded file" for
computer processing.

In the processing of any large file for computer
use and pre-selection, screening of documents, ab-
stracting, or indexing whereby any portion of the
information is omitted, condensed, transformed or
classified by human minds prior to its recording
in computer-usable form will invariably lead to cor-
responding errors and omissions in retrieved or re-
trievable subject matter. As mentioned before,
no two human minds can be reliably counted upon
consistently to make the same decisions as to what
material is to be included or as to the words to be
selected for the abstract or index. It is far better,

212

(within economic boundaries provided by the law of
diminishing returns) to record initially the entire
text of all documents conceivably relating to the
general subject matter desired to be searched.

The computer can then be programmed reliably
to generate sub-classes or abstracts or indices in
accord with programmed definitions and intentional
margins for error adjusted in accord with empirical
evidence. The excluded subject matter will thus
be rediscoverable by inspecting the definition or pro-
gram, any errors correctible, and the initial material
still ultimately retrievable from other viewpoints, if
desired.

Secondly, there must be procedures for direct re-
trieval at several different levels of "idea complex-
ity". The differing nature of each file as well as
the differing nature of and purpose to be served
by the information to be retrieved requires great
versatility in the manner in which the search ques-
tion may be framed.

In one case, for example, it may be logical to
search for all documents in which a given single word
( representing a single concept ) appears. In another
case, it may be more logical to search for all sen-
tences in which a certain phrase (representing a
"complex concept") occurs. In yet another case,
a search for all paragraphs containing a selected
combination of key words or phrases (representing
a "complex idea") may be indicated. Any large
searching center should thus be able to accommodate
all of these as well as a host of other possible com-
binations of specificity and genericity in the language
and structure of the search.

Thirdly, there must be procedures within the com-
puter for relating words and phrases of similar mean-
ing to one another (thesaurus) as well as for relating
complex-concept words to combinations of more
rudimentary concept words which express similar
meanings ( semantic factoring ) .

For reliability in retrieval, it is necessary to nor-
malize the language involved. The same idea may
often be expressed by several hundred different com-
binations of words. In the ideal situation, the
computer should be able (through language nor-
malization techniques) to search for a given idea
regardless of its various possible modes of expression
in any document so long as the idea to be searched
is submitted to the computer in accord with one of
the possible modes of expression.

Here again, it is virtually impossible for different
human minds reliably and consistently to accomp-
lish this language normalization function even with

external aids. Because of its complexity, this func-
tion must be done by the computer itself.

Fourthly, there must be procedures for determin-
ing and adjusting retrieval on the basis of the per-
tinency of the ideas as expressed in the various
documents being searched to the idea as expressed in
the search question.

In most cases, a searcher wishes to retrieve not
only all documents which completely express the
particular idea being searched (total coincidence
with the terms of the search question ) but also those
documents which express similar ideas but not pre-
cisely the same idea. Moreover, experience has in-
dicated that it is virtually impossible to frame a
search question with sufficient accuracy and com-
pleteness that all possible variations of expression of
a particular idea will be retrieved if all of the condi-
tions of the search question must be met in the ex-
pression of the document.

By permitting an adjustable degree of leeway in
the pertinency of the document to the question be-
ing asked, a predetermined amount of "overpull"
can be achieved which will greatly increase the re-
liability of retrieval. As the requisite degree of per-
tinency is relaxed, a greater volume of documents is
retrieved and reliability improves but the number
of non-pertinent documents retrieved (false drops)
also correspondingly increases. However, with any
given type of file (i.e., legal, patent, medical, etc.),
experience eventually indicates the pertinency fac-
tors which should be used with various types of
search questions in order to obtain the best com-
promise between reliability and resolving power.
Moreover, as the language normalization techniques,
computer programs and question-framing tech-
niques are improved and perfected, readjustment to
ever higher degrees of pertinency will be possible
until excellent reliability is eventually achieved at
the sacrifice of comparatively few false drops.

Fifth, there must be procedures for "weighting"
the various concepts (words or phrases) of the
search question in accord with the relative impor-
tance of these concepts to the content of the search.

Particular words or phrases expressing one of the
component concepts of an idea being searched might
have far more significance to the searcher than other
words or phrases expressing other component con-
cepts.

For example, in a search for all patents disclosing
semiconductor oscillator circuits having capacitive
feedback, the concept "semiconductor" might need
to be weighted more heavily than the concept "oscil-

213

lator" if the searcher considered that he had more
interest in "semiconductor" circuits with capacitive
feedback than in "oscillator" circuits with capacitive
feedback. This also becomes important in the
adjustment of pertinency factors. For example, in
this search for a semiconductor oscillator circuit, the
searcher might be willing to accept a pertinency
of 80% between document text and search ques-
tion if the text included the term "controlled-recti-
fier" or equivalent, but might require a pertinency
of 90% if this term were not present.

Sixth, there must be procedures for statistical
analysis of the words both in the total file as well as
in the individual documents. This is helpful both in
selecting the words of the search question and in pro-
gramming the search itself. For example, the rela-
tive frequency of word occurrence is a good measure
of its discriminatory value. Moreover, the number
of times a given concept appears in a given docu-
ment is some measure of the emphasis which the
document places upon the particular concept in ques-
tion, and thus may help decide pertinency.

Seventh, there must be procedures for determining
the order of and relationship between words (syn-
tax, role-indicators, etc.).

In some searches because of contiguity in meaning
of a great many documents in a particular file, it
becomes necessary to distinguish between the "dog
bites man", and "man bites dog" situations. This
tends to be particularly true where the search ques-
tion requires the coincidence of only a relatively
few (for example, less than 5 ) simple words or con-
cepts. As the number and meaningful complexity
of the words in the search question increases, the
necessity to monitor the order of and relationship
between words becomes less urgent since the prob-
ability that two different ideas will occur in a given
file which contains the same combination of, for ex-
ample, ten words in different order is very small.
For example, while three words may have different
meanings if in different relationship to one another,
it is quite unlikely that both of these three-word re-
lations will make sense when combined with seven
other words to express a complex "idea". Neverthe-
less, in order to be able to handle all types of search
requests, it is highly desirable that the computer
search program of any large search center be ver-
satile enough that particular relationships between
selected words or concepts may be required to sub-
sist before the text or document will be retrieved.

And finally, there must be relatively routine pro-
cedures for human and machine modification and

improvement in accord with the results of actual re-
trievals of the stored and programmed semantic
codes and their relationships.

In other words, there must be feedback refine-
ment of the coding and programming. The reasons
for this requirement are believed self-evident.

So much for the requirements of the system.

As far as the feasibility of this program is con-
cerned, I would like to call attention to the activities
of the University of Pittsburgh Health Law Center,
under the direction of John Horty, which has
already taken the first steps toward the development
of a mechanized legal searching system which may
ultimately fulfill all of the basic requirements for
large searching centers enumerated above. At this
Center they have already succeeded in transcribing
the entire text of the Statutes of Pennsylvania onto
magnetic tape. Each word is addressed in computer
storage by its location in each sentence of each
statute. An alphabetical vocabulary listing of all
words in the statutes as well as their respective fre-
quencies of occurrence has also been generated.

Searches are conducted on the basis of different
combinations of words, selected by human minds
from the vocabulary list, that are most likely to be
found in statutes relating to the idea being searched.
An attempt is made to include all synonyms of the
words in each combination and the computer is pro-
grammed to identify all statutes in which the re-
spective word combinations occur. The locations of
selected words relative to one another can also be
programmed during the search; for example, the
words "room" and "board" can be programmed
in the search question to occur within two or possibly
three words of one another before the statute is
retrieved.

It will thus be seen that this University of Pitts-
burgh activity already meets several of the docu-
mentational and operational requirements for large
searching centers postulated above. First of all,
since the full text of the statutes is read into storage,
it meets the requirement for "comprehensiveness"
in the creation of the initial master recorded file.
Secondly, since a great variety in the number or
combinations of different words in either sentences,
paragraphs or entire statutes may be called for in
the various search questions, it meets the second re-
quirement of procedures for direct retrieval at sev-
eral different levels of "idea complexity". Finally,
it partially meets the requirement of procedures for
determining the order of and relationship between
words.

214

The requirements which remain to be met relate
to the development of computerized procedures for
searching through a file-generated thesaurus which
includes the word linkages resulting from semantic
factoring, as well as the generation of computer pro-
grams for adjustable pertinency and weighting fac-
tors which take advantage of prior statistical analysis
of the frequency and discriminatory value of the var-
ious words in the total file.

Mr. Horty and his associates at the University
have already embarked upon the task of creating
a file-generated thesaurus. In conversations with
them during a recent visit to the University, a tenta-
tive long-range step-by-step program for meeting the
remaining requirements was laid out.

Now let me take a few moments to talk specifically
with respect to the patent searching problems of the
United States Patent Office.

It is well recognized that the mechanization of
patent searching in the Patent Office must necessar-
ily take several years. During this time, I believe
we can expect the introduction of efficient optical
scanners capable of relatively inexpensively and
automatically reading the full text of a patent into
storage, as well as the introduction of new com-
puters with substantially greater storage capacity
and more rapid access time. This will, for reasons
both of economy and reliability of search, ultimately
lead to the introduction of the full text of the entire
or designated portions of each patent into storage
thereby eliminating the need for two independent
but difficult-to-harmonize human translations. The
hopeful aspect of this likely evolution is that the
experience currently being achieved in developing
systems for classifying and retrieving in accord with
coded indices will be useful in establishing the codes
to be used by the machine rather than by the human
mind in retrieving directly from the full text. In
short, when the optical scanner finally arrives, we
hope to be in a position to take advantage of it.

However, we must also begin to build up some
actual experience in mechanized procedures for re-
trieving directly from the full text. For example,
it is believed that reliability of technical information
retrieval cannot be achieved with good resolving
power at greater than the complex idea (para-
graph) level, but this should be verified by experi-
mentation. If it is true, each complex idea
(paragraph or, in some cases, sentence) of the pat-
ent will need to be given a separate address in
storage and to be independently retrievable.

Other questions to be answered are: how to form
the search question, and how to program the com-
puter with its stored glossaries, translations, syno-
nyms, factored words and compiled words and
phrases so that only patents of controllable perti-
nency will be retrieved.

In my opinion, the trained patent searchers of
the future will merely express the search questions
in precise natural language, and the computer will
take each meaningful word in the query, factor it
into its component meanings, recombine the various
factored words in to all synonyms and complex
words which convey identical or similar meanings
and match these recombined alternatives against
the text of the documents in accord with pro-
grammed syntactic relations and against a level of
pertinency established by the searcher!

However, the starting point must be a rigorous
analysis of the current mental gymnastics of a
trained searcher as he reviews the full text of pat-
ents being searched. The building of technical
glossaries which take into account the basic concepts
inherent in complex words and ideas (and the vari-
ous combinations of such concepts by which similar
complex ideas may be conveyed) must also soon
be undertaken somewhere. These activities should
be undertaken and measured against a backdrop
of actual experience in formulating retrieval ques-
tions and comparing the results against previous
manual searches. They might, well be tackled by
the Universities in cooperation with the Patent Of-
fice. As another immediate step the full text of
each newly issuing patent might be preserved in
machine readable form by being automatically
punched into paper tape as the type for printing the
patent is set.

As yet another immediate step the full text of
the patents in one relatively new but potentially
important field of art, for example, cryogenics, may
be placed onto magnetic tape, and experiments in
retrieval from the full text begun. The experience
resulting from such experiments would be invalu-
able as the computer programs for full text search-
ing of larger and larger segments of the patent file
are eventually developed.

From an economic standpoint, society can best
afford this investigation where the document to be
retrieved is per se valuable. It will be difficult to
find any file of documents in our society other than
patents where the document itself is so ideal for the
experiment. A patent is a precisely written docu-

645803 O-63-10

215

ment with little surplusage. Each sentence is ordi-
narily loaded with meaning. It has intrinsic value
to society because of its informational content.
Moreover, it has intrinsic value to society and to
its owner as a property right.

For all these reasons, it is believed that industry,
the public and the various Bar Associations should
all support a greater effort toward the institution
of investigations in the Patent Office with respect to
the theory and procedures for searching the full text
of patents upon the basis of machine processed basic
language -type semantic codes. Such a program
will also hold promise for the solution of some of
the more subtle problems involved in mechanized
searching of more general legal resource materials,
such as judicial decisions.

Ladies and Gentlemen, the time for polite and
disconsolate hand-wringing over insurmountable
problems has passed. We must, as a nation and as a
profession, begin to reach some common agreement
as to a program for action if the rapid progress of
our industrialized society is to continue. While the
leadership and decisions must necessarily come from
the United States Patent Office itself, the American
Bar Association and the American Patent Law As-
sociation can be helpful in marshalling the resources
and cooperation of the Bar in the furtherance of an
agreed-upon program. The alternative is the slow
strangulation of our patent system and a consequent
erosion of private initiative. Let us not hesitate to
speak out boldly and to act imaginatively and
promptly.

216

APPENDIX AA

THE PATENT SYSTEM AND THE TECHNOLOGICAL REVOLUTION

Dr. Alan T. Waterman
Director, National Science Foundation

Presented at the Celebration of the American Patent System, Sheraton Park Hotel,

Washington, D.C., October 19, 1961

Dr. Alan T. Waterman: Mr. Commissioner, Dr.
Bush, distinguished guests, ladies and gentlemen,
it is a privilege to be with you this evening on one
of the several occasions that have been arranged
in celebration of the 125th Anniversary of the
Patent Act of 1836. Having only just observed the
10th Anniversary of the agency with which I am
associated, I am awed by such venerability. I hope
that when the National Science Foundation attains
such ripe old age, the occasion will be marked by
as distinguished a company as is gathered here
tonight.

When I asked myself what point of view I, as a
relative outsider, should present this evening, I re-
flected that probably there is greater knowledge of
the patent system concentrated under this roof than
has ever been assembled before in one place. There
would be nothing I could add, therefore, to what
you already know about this important aspect of
our economic system. Furthermore, the National
Science Foundation is primarily interested in basic
research, which in its own domain, is antithetical to
the purposes for which the patent system was estab-
lished. It is the common belief of scientists generally
that new knowledge and discoveries of the working
of nature's laws should be freely available to all with
no restrictions of any kind.

Several years ago, in describing the patent poli-
cies of the National Science Foundation to one of
the Senate subcommittees, I observed: "The nature
of basic research is such that the possibility of pat-
entable inventions resulting during the conduct of
such research is very limited. It has been stated
from time to time in judicial decisions relating to
patent matters that a conception of the mind, not
represented in some physical form, is not an inven-
tion within the meaning of the patent laws and that
patents are not granted on discoveries of abstract
laws or principles that can be awarded only with
respect to the embodiment of the concrete idea,
whether such embodiment is a device, a product or
a process. Since basic research is concerned pri-
marily with the study of natural phenomena and
the ascertainment of natural laws and principles,
the prospects of patentable material being contained
in such research are not very substantial, particularly
patents having commercial value. I hope you won't
hold it against us that we are not too much involved
in patents. We do appreciate their importance
nevertheless.

Despite its emphasis on basic research, however,
the Foundation has subsidiary responsibilities which

I believe are relevant to a discussion of the patent
system. One of these is the collection and analysis
of data relating to the nature and growth of the
national research and development effort; and
another is a continuing concern with problems re-
lating to the more effective storage and dissemina-
tion of scientific information. It seemed to me,
therefore, that we might properly ask whether the
patent system in its present form is keeping pace
with the progress of research and development, and
whether we might profitably apply to the operations
of the Patent Office itself some of the new tools that
technology makes available.

It is fashionable nowadays to relate everything to
the technological revolution; hence, I have chosen
as my topic, "The Patent System and the Tech-
nological Revolution."

We Americans are fond of subjecting to endless
analysis and discussion even our most highly re-
spected institutions. Obviously the world in which
we live today is much further removed from the
world of 1790 than the mere lapse of time would
indicate. For example, we are told that more sci-
entists and engineers are living today than in all
other ages combined. It is all the more remarkable
then that in the still largely agrarian economy the
Founding Fathers should have had the foresight to
provide for the establishment of a patent system that
would stimulate and encourage growth of an in-
dustrial society. We salute, also, the orderly in-
stincts that prompted the revision of the patent sys-
tem as early as 1836 and the passage of the Act that
charted the course of its operations for the next
century and a quarter. In the last two decades,
however, there have been many increments of
growth in the economy as a whole; and research and
development, to which the patent system is so
closely linked, have multiplied at a rate much faster
than at any previous time.

The National Science Foundation, in a program
of comprehensive studies made on a periodic basis,
attempts to measure various aspects of the nation's
scientific effort. Beginning with the year 1953, these
studies have yielded trend data on the total number
of scientists and engineers engaged in research and
development and on expenditures for the perform-
ance of research and development by industry, the
Federal Government, colleges and universities and
other non-profit institutions. Data on science and
engineering in the United States for the years prior to
1953 are fragmentary but can nevertheless be used
for the purpose of drawing broad comparisons be-

219

tween present and past levels of research and devel-
opment activity in our economy.

The growth and importance of the research and
development activities of scientists and engineers
during the last 30 years is demonstrated by the nearly
eight-fold increase in the numbers of scientists and
engineers engaged in such work — from 50,000 in
1930 to 395,000 in 1960. The most spectacular
measures of the growth of research and development
during this period, however, are those relating to
expenditures. The best available estimates indicate
that less than $200 million was spent for research
and development in the entire country in 1930, com-
pared with 1960 outlays totaling approximately $14
billion.

In contrast to the tremendous increase in the vol-
ume of R & D activities characterizing the Post-
World War II period, the numbers of patent appli-
cations and of patents granted for inventions have
not varied appreciably during the period. The num-
ber of patent applications has increased from the
Post-World War II low of 65,000 in 1951 to a high
of 84,000 in 1960 . It is interesting to note, however,
that the number of parent applications reached a
peak during the 1920's. During the ten-year period
ended in 1930, patent applications averaged 89,000
per year compared with annual averages of 70,000
during 1931-40, 69,000 during 1941-50, and
79,000 during 1951-61. The number of patents
issued for inventions was highest in 1932. In this
connection, one might wonder whether or not there
is any relationship between depressed economic
conditions and inventive ingenuity. During the
ten-year period ended in 1940, patents issued for
inventions averaged 44,000 per year, compared
with 31,000 per year during 1941-50 and 43,000
per year during 1951-60.

One must approach with caution, however, the
seemingly anomalous situation in which tremendous
increases in scientific research and development are
accompanied by relatively small increases in the
number of patents granted for inventions. During
recent decades, patented inventions have increased
enormously in scope and complexity. There are
more comprehensive applications from large firms
and fewer from individuals.

It is a pity, incidentally, that funds have not been
provided for a patent museum in which the models
for patents that were once required to accompany
applications could be publicly displayed, and an even
greater pity that many of these models were com-
pletely destroyed in a disastrous fire. Otherwise, one

220

might be able to see with his own eyes how American
inventive genius has advanced from the simple to
the complex.

In the "attic" of the Patent Office, a few of these
models survive. There is a model of a "corpse
cooler," for example, consisting of a wicker table of
somewhat macabre shape; there is a walking cane
embellished with a coiled serpent, the American Flag
and Holy Bible and more elaborate models of steam
engines and early reapers. When one compares
these objects with the electronic data system for
processing bank checks at a speed of 750 per minute,
for which the 3,000,000th patent was issued in
September, one gains some notion of the progress
of technology. There has been a shift, furthermore,
from the individual inventor working on his own,
to the corporate inventor working as a member of a
team of specialists with all the resources of a modern
industrial laboratory at their command. I am told
that today about 70 per cent of all patents are issued
to corporations.

One must also take into account the role of the
Federal Government. More than half of the esti-
mated $14 billion expenditure for research and de-
velopment in the United States is financed by the
Federal Government. Much of the federally-
financed research and development work is allocated
in the fields of aircraft and missiles, nuclear energy
and other defense-related fields, and security classi-
fication becomes a factor in the number of patent-
able inventions.

The preponderance of Federal funds in the na-
tional research and development effort also raises
the point, which is periodically debated, of whether
or not the Government should take title to patents
that are issued under its research and development
contracts. I do not wish to take sides in a con-
troversy in which a vast amount of information and
talent have been marshalled to support each point
of view. I do wish, however, to emphasize the
need for a wise policy with respect to the administra-
tion of such patents as may arise from Government-
sponsored research and development. Such a
policy should take into account the fact the patent
system was designed to promote the use of new in-
ventions, and unless inventions are used, the public
interest is not served. Obviously, the performance
of Government-sponsored research on problems re-
lating to defense, space, and atomic energy, pro-
duces incidentally new products and processes oi
great commercial value. A policy governing their
use should preserve industrial incentives to partici-

pate in Government enterprises and insure that
scientific information is suitably disseminated. It
might be well to consider the establishment, per-
haps as a unit in an existing agency, of a central
Federal inventions organization which would have
the responsibility of devising suitable licensing pro-
cedures, and seeing that Government patents are
put into use.

To return to the relationship of research and de-
velopment to patentable inventions, I should like to
discuss one or two additional factors that help to
explain the apparent lack of correspondence be-
tween the growth of research and development and
the issuance of patents. In recent years, about two-
thirds of the nation's research and development ex-
penditures have been allocated to development.
Development includes technical activities of a non-
routine nature concerned with translating research
findings or other scientific knowledge into products
or processes. Development frequently involves the
utilization of large numbers of scientific, technical
and other supporting personnel and extensive mate-
rials in the building and testing of models or proto-
types. A considerable portion of Federal R&D
expenditures is for development in such fields as
missiles, atomic energy devices, space exploration
vehicles and equipment, and electronic guidance
and control systems. It is likely, therefore, that
relatively fewer inventions per dollar expended re-
sult from the development phase of R & D projects
than from applied research.

As I mentioned previously, that portion of the
nation's R&D expenditures allocated to basic re-
search performance results in comparatively few
patentable inventions. In 1960, funds for basic
research performance totaled slightly less than 1.5
billion (including capital facilities) or approximate-
ly one-tenth of the nation's total research and de-
velopment effort measured in terms of dollars.

Although the rate of patent applications and
patents issued cannot be regarded as a true index of
the growth of research and development, neverthe-
less, I believe it is pertinent to explore the question
whether or not the patent system in its present form
is going to be equal to the anticipated growth of
science and technology.

The National Science Foundation has recently
completed a ten-year projection that predicts a
steady rise in the number of scientists and engineers
and in dollar expenditures for research over the next
decade. The number of doctorates in science and
engineering, for example, is expected to be approxi-

mately double the present number by 1970. The
U.S. Patent Office, meanwhile, is operating under
much the same systems and procedures that it did
fifty years ago. The patent applications have grown
in size and complexity and the material to be
searched has doubled and redoubled, so that the
time required for each search grows commensur-
ately longer. In an age that is characterized on
every side by a high degree of mechanization and
automation, the individual examiner is for the most
part obliged to conduct his searches by what is es-
sentially a primitive manual method. It is reason-
able to expect that the advances of science and
technology can be applied to this part of the process
so that the bulk of the examiner's time can be de-
voted to the intellectual aspects of his tasks. A
mechanized search procedure, properly conceived
and executed, would materially reduce the time
required for each search, would eliminate the factor
of human fatigue and could be expected to "discover
remote and previously unobserved connections,"
thus contributing to the issuance of better patents.

Such a proposal is, of course, by no means new.
The problem has been examined and re-examined a
number of times and the Patent Office has a re-
search program directed toward the design, con-
struction, and experimental use of mechanized
search systems, although it has had neither the funds
nor the numbers of personnel for an effort commen-
surate with the magnitude of the problem.

In 1954 I had the pleasure of serving on an Ad-
visory Committee, of which Dr. Bush was Chair-
man, which examined the whole question of the
application of machines to Patent Office opera-
tions. The problem has two major aspects: the
physical capacity of existing automatic data proc-
essing equipment for storing the vast amount of in-
formation involved, and the intellectual task of
reducing the information to a form that can be
handled by machines.

On the basis of certain arbitrary assumptions re-
garding the size of each patent, the Bush Commit-
tee decided that for the 1954 file of five million
foreign and American patents, a mechanical stor-
age capacity would be required for at least 300
billion bits. A "bit," as you know, is a technical
term in information theory used to designate a unit
of information. Allowance would have to be made
for annual accretions, again excluding drawings,
at the annual rate of 3,200,000 bits. Storage capac-
ity required for coding indexed entries and other
selections of criteria was estimated at around 50

221

billion bits for the patent files in 1954. Accretions
since 1954 have raised the total number of United
States and foreign patents to 8.5 million and the
related scientific and technical literature has in-
creased by a much higher factor.

Since comme"rcially available reels of magnetic
tape can store up to 100 million bits of information,
the 300 billion bits estimated by the committee
could be stored on 3,000 of these reels. There are
digital computer installations that have libraries of
several thousand reels of tape; hence, this amount of
storage is now available. New developments give
promise of achieving the same amount of storage at
less cost in smaller size and with more rapid access.

Having determined that the storage capacity
which would be required of an automatic indexing
and retrieval system for the Patent Office was realiz-
able, the Bush Committee went on to note: "The
creation of a classification and indexing system so
that machinery can be effectively used for searching
is an intellectual problem of formidable proportions
because it involves a diversity of complex subject-
matter fields." The Committee recommended an
intensive program of research and development
within the Patent Office working toward the adapta-
tion of machines to its specific and unique problems.

In the six years since the Bush report was sub-
mitted, the Patent Office has established a mecha-
nized system for the searching of steroid patents and
similar work is under way in the field of polymers,
chemical processes, organic phosphates and elec-
tronic arts. It has logically begun with the arts
that are simplest to adapt to machine processing,
on the theory that as the work progressed, rapid
advances in the development of automatic data
processing equipment would provide increasingly
sophisticated equipment for the more complex
subject matter.

We congratulate the Patent Office on the excellent
work that it is doing in this area, but I believe it is
fair to say that a greatly expanded research and
development activity is needed if the work of mech-
anizing the search procedure is not to be outstripped
by the rapid accretions of the material to be searched.
This expanded research program should, of course,
be closely related to other current work being done
in the broad field of information retrieval.

The National Science Foundation has a special
interest in the Patent Office problem because it
represents in microcosm the whole scientific infor-
mation problem for which the National Science
Foundation has direct and continuing concern.

The volume of scientific information is increasing
at an exponential rate which makes it impossible for
the individual scientist to keep abreast of the pub-
lished material in his own field and in related fields
in which he may be interested. There are great
problems of storage, retrieval, dissemination and
translation from one language to another.

An increasing amount of research is being de-
voted to these problems, which is regularly reported
by the Foundation in a semi-annual report entitled
"Current Research and Development in Scientific
Documentation". This research is primarily con-
cerned with the handling of information in lan-
guage and other non-quantitative forms such as
photographs and circuit diagrams as distinguished
from quantitative forms or numerical data. The
latter can be quite readily coded, stored, and
manipulated by widely understood electronic data
processing techniques. The mechanized handling
of ideas, concepts, and techniques embodied in texts
and drawings present much more difficult problems
and these are yet to be resolved.

To illustrate the scope as well as the interest of
some of the work that is being done in documen-
tation research, I cite a few examples. The
Ramo-Wooldridge Division of Thompson-Ramo-
Wooldridge, Inc., and the International Business
Machines Corporation are attempting to devise
automatic indexing and abstracting techniques.
Research groups at the University of Pennsylvania
and the Radio Corporation of America are con-
centrating on the mechanization of certain tech-
niques of linguistic analysis of English, with the
thought that such analysis will prove to be an essen-
tial step in automatic indexing, abstracting, and
retrieval. The National Bureau of Standards is
engaged in research on the mechanized processing
of both language-texts and pictorial and other
forms of graphic information. Ideally, an infor-
mation processing system should be able to handle
both types of information, for they are usually
encountered together in scientific papers and in
patents.

Western Reserve University has developed par-
tially mechanized information retrieval procedures
which are being tested in the field of metallurgy.
A group at the Itek Corporation is doing research on
a special documentation language based on logical
principles and a systematized technique for indexing
the concepts and relationships in scientific papers.
The British Association of Special Libraries and In-
formation Bureaux is in the third year of a con-

222

trolled comparative study of the searching effective-
ness of several indexing and classification systems.
The Chemical Abstract Service, using existing equip-
ment and techniques, is experimenting with ways
of coding data on the structure and properties of
chemical compounds for mechanized storage and
retrieval and for automatic preparation of data
compilation.

In the field of mechanical translation, about a
dozen research groups in the United States are pre-
paring automatic dictionary machine programs for
a number of scientific fields, devising machine pro-
grams for the grammatical analysis of certain
foreign languages, working on techniques of se-
mantic analysis and on the synthesis of English
sentences — and, in general, are gaining deeper in-
sights into the nature of language.

Although the foregoing gives only a hint of the
present scope and objectives of research on informa-
tion retrieval, a great deal more remains to be done
before we know how best to use machines to aid the
handling of information and patent searching.

Nevertheless, the expected consequences of such
research are of the greatest importance for science
and industry. Improvements in the means for con-
sulting the records of accumulated scientific knowl-
edge will certainly benefit researching of the patent
records.

In the meanwhile, the backlog of unsearched
patents remains constant, the elapsed interval of
time for the granting of a patent has increased and
the over-all number of items in the patent file has
almost doubled. A little over a year ago, IBM esti-
mated that industry was spending about $2 million
a year on systems for information retrieval and that
by 1965 the expenditure will jump to over $100 mil-
lion a year and will double every three years
thereafter.

The Patent Office has been among the pioneers
in research on the use of mechanized search tech-
niques, but like others active in this field, has not
yet been able to devise ways of using to the fullest
the capabilities of the remarkable electronic devices
produced by our advanced technology. If the
Patent Office were to be given the funds and the
staff to tackle this problem on the broad scale war-
ranted by its importance, it would not only enhance
the efficiency of the patent system but it would un-
doubtedly contribute in a major way to the solution
of scientific information problems which are of
pressing importance to industrial research, and in
fact to the scientific community as a whole.

Before closing, I should like to say a few words
about even broader needs for research and data re-
lating to the patent system. Although we are here
to pay tribute to the patent system and to contribu-
tions it has made to progress in the last century and
a half, we cannot close our eyes to the fact that the
system is frequently subject to attack and the sug-
gestion has been advanced from time to time that
it might be eliminated entirely.

In June of this year, The Division of Engineering
and Industrial Research of the National Academy of
Sciences-National Research Council sponsored a
symposium on The Effect of Patents on Research.
Members of the symposium were drawn from big
business, small business, the universities, economics
and patent law. During two days of intensive delib-
erations, these people who are well informed on the
subject of patents, contributed their experience and
point of view to discussion of the problem. The
report of this committee has not yet been published
and I do not wish to anticipate its findings. I think
I may say, however, that the group agreed that al-
though the merits of the present system have been
debated pro and con, careful study is required before
changes other than administrative or procedural
should be considered. The group further agreed
that our nation's technical progress could be ser-
iously affected by ill-considered action that might
disrupt present incentives and methods of informa-
tion dissemination.

This conclusion emphasizes, in my opinion, the
need for much more research and study on the pat-
ent system as a whole and its effects upon the econ-
omy. To date relatively little has been done in this
field. The rapid advance of computed technology
has made possible, among other things, the construc-
tion of models for the study of complex, socio-
economic situations. Through its Social Sciences
Division, the National Science Foundation is sup-
porting several studies on the economic aspects of
inventions and on the economic implications of re-
search and development. In addition, we have a
specific contract for the study of factors affecting the
diffusion of innovations in industry. May I say that
the Foundation would welcome additional sound
proposals in these fields where clearly much study
remains to be done.

It is my personal opinion that by and large the
patent system has served us well and it has been a
considerable factor in the growth of the inventive-
ness and ingenuity for which as a nation we are
famous. We are, however, in the midst of a tech-

223

nological revolution in which the character of in-
vention seems to be changing rapidly. Whether
the patent system as we now know it can continue
to serve adequately the needs of an expanding
research and development effort and enable us to
meet foreign competition successfully is a question
that cannot be answered because we lack sufficient
data.

In any event, it is timely to apply to an assessment
of the patent system the orderly processes of re-
search, and by all means to place on the most mod-

ern basis possible the techniques of patent processing
so that we may enjoy the maximum benefits of a
system that has long been a bulwark of free enter-
price. ( Standing ovation. )

Commissioner Ladd: Dr. Waterman, we are most
grateful to you for your thoughtful, penetrating
discussion.

It has been a wonderful evening, good-night to
you all. (Applause.)

(Whereupon, at ten twenty o'clock p.m., the
meeting adjourned.)

224

APPENDIX B

Summary of Answers to Questionnaire

The patent offices of the countries invited to at-
tend the International Patent Office Workshop on
Information Retrieval were requested to answer a
questionnaire directed to patent classification and
search facilities and the material searched when
examining a patent application. Replies were re-
ceived from the patent offices of Australia, Aus-
tria, Canada, Denmark, Germany, Japan, The
Netherlands, Norway, Sweden, the United King-
dom, the United States and from the International
Patents Institute.

The attached pages contain a summary of the
answers to the questionnaire, prepared by W. S.
Cole, Assistant Primary Examiner, Division 94,
U.S. Patent Office.

1 . Which of the following criteria are used as a
basis for a patent grant? A. Novelty; B. Utility;
C. Invention.

Summary: Novelty, utility and invention are all
essential to patent grant in each of Austria, Canada,
Denmark, Germany, Japan, the Netherlands, Nor-
way, Sweden and the U.S.A. In Australia and the
United Kingdom, novelty and invention are the only
requirements but, procedurally, in the United King-
dom the question of lack of invention is raised, if at
all, only at the opposition stage (after publication but
before grant ) . The question of basis for grant of
a patent is of course not relevant in the case of the
International Patent Institute which does not issue
patents.

2. Which of the following types of literature are
searched in determining patentability? A. Na-
tional patent literature; B. National technical liter-
ature other than patents; C. World wide patent and
other technical literature.

Summary: During the examination of an appli-
cation for patent the available patents and technical
literature of the entire world are searched by Aus-
tria, Germany, Japan, Norway, and the U.S.A.
The International Patent Institute conducts searches
on a similar scale to determine novelty of disclosure
or validity of patents. The United Kingdom and
Australia search their own "patent literature" (i.e.

patents and patent abstracts) but while the search
of patents of other countries and of other literature
is discretionary, it is apparently not usually under-
taken. The Netherlands search includes patents of
Germany, United Kingdom, France, U.S.A.,
Switzerland, Belgium, Luxembourg and the Nether-
lands and other literature to a limited extent. Ca-
nada and Denmark search their own patents and
technical literature and to a limited extent that of
other countries; Denmark's search includes patents
of Norway, Sweden, Finland, Germany, France, the
United Kingdom, and the U.S.A. Sweden's search
is compulsory through the patents of Sweden, Den-
mark, Finland, France, Great Britain, Norway,
Switzerland, Germany and U.S.A.

3. Is the search made A. Without regard to the
age of the literature? B. If there is an age limita-
tion on the literature which is searched, state the
limitation.

Summary : Germany does not search patent and
technical literature which is over one hundred years
old. In the United Kingdom and in Australia, for-
eign and domestic patents which are more than fifty
years old and official extracts thereof may not be
cited. However, in the United Kingdom techni-
cal materials other than patents may be cited re-
gardless of age. Austria, Canada, Denmark, Inter-
national Patent Institute, Japan, the Netherlands,
Norway, Sweden, and the U.S.A. may employ avail-
able domestic or foreign patents and technical liter-
ature regardless of age as evidence of unpatenta-
bility.

4. Indicate which of the following are permis-
sible subjects for a patent grant. A. Manufactur-
ing process; B. Non-manufacturing process; C.
Machine; D. Article of manufacture; E. Static
structure; F. Materials from which articles are
manufactured; G. Chemical compounds; H. Ad-
mixtures of chemical compounds with other sub-
stances.

Summary: The different types of inventive sub-
ject matter upon which a patent may be granted
include; (A and B) Manufacturing and Non-

225

Manufacturing processes, (C) Machines, (D)
Articles of Manufacture, (E) Static structures and
(H) Admixtures of Chemical compounds with
other substances in Australia, Canada, Germany,
Denmark, Japan, the Netherlands, Norway,
Sweden, the United Kingdom and the U.SA.
However, while Canada may allow certain types of
non-manufacturing processes which involve, for
example, prospecting, mechanical analysis and elec-
trical communication, other non-manufacturing
processes such as medical or surgical procedures,
agricultural methods, chemical analysis and the ap-
plication of pesticides and the like are not allowed.
Some agricultural, medical and biological processes
are also excluded in the United Kingdom. In the
Netherlands new materials are not patentable per
se but may be protected by a patent claiming the
method of making the material even where this
method is not inventive. A patent is permissible
for chemical compounds admixed with other sub-
stances which have some shape, e.g., fibres, flock,
webs, yarn. Mere chemical admixtures fall under
the practice outlined under (G) below. The
United Kingdom, Japan and Australia refuse
chemical mixtures which consist of foods and medi-
cines whose ingredients act in concert in the same
manner that they would act individually. This is
also true for Canada where the principle extends
to all mixtures and parts of machines and is called
"mere aggregation." In the U.S.A. therapeutic
or medicinal inventions must have proved efficacy
to be patentable. Austria grants patents for manu-
facturing and non-manufacturing processes, ma-
chines, articles of manufacture and static structures
but not for non-manufactured (i.e. raw) materials,
chemical compounds or admixtures thereof. With
respect to Item (F), materials from which articles
are manufactured may form the basis of a patent
grant in Australia, Canada, Denmark, Germany,
Japan, Norway, Sweden, the United Kingdom and
the U.S.A. In Germany and Norway alloys may
be patented and in the United Kingdom, any new
substance which is a product of manufacture such
as an alloy, food or drug may be patented. In the
two preceding cases, the alloys may be regarded as
a species of Item (F) (i.e. materials from which
articles are manufactured ) since this is the usual use
of an alloy. In Germany, the application of a
known material to produce a known article is not
patentable if it represents a mere replacement with-
out the creation of a new effect. As to Item (G)
claims to chemical compounds may be allowed in

226

Australia, Canada, the United Kingdom and the
U.S.A. In the Netherlands, in Germany and in
Norway the raw material or chemical compound
can be protected only by means of a claim reciting
its method of manufacture, as stated before. Den-
mark does not allow claims to a chemical com-
pound, but to the use of the compound. In
addition, foodstuffs, table luxuries and pharmaceu-
ticals are often not patentable in Germany. Mis-
cellaneous types of subject matter are considered
appropriate for a patent grant as follows: In
Canada, viruses and vaccines may be patented
but not plants or bacteria. As to antibiotics,
Norway grants patents for manufacturing proc-
esses only. The U.K. and the U.S.A. also grant
patents for antibiotics. In the U.K., methods of
test applicable to the control of manufacture are
patentable. With respect to Items F, G and H,
however, a patent may not be granted in the U.S.A.
for a raw material or chemical compound or ad-
mixture which is a naturally occurring substance,
and in the U.K. patents for new substances do not
extend to the substances when found in nature.
The U.S.A. may grant patents for ornamental de-
signs based solely upon appearance and patents for
botanical discoveries and inventions in the nature
of asexually produced new plants, provided that
they are not tuber propagated and are not found in
an uncultivated state. Bacteria are not patentable
in the U.S.A.

5. When searching, which of the following cri-
teria are used to determine the relevance of reference
documents? A. Novelty is the sole consideration;
B. Analogous arts are considered; C. A doctrine of
equivalents is relied upon.

Summary: In part (A) of this question, it must
be remembered that part (A) is closely related to
parts ( B ) and ( C ) . That is to say, in part ( B ) ,
analogous arts are searched where known uses of
substantially the same concept as that sought would
be considered a valid reference and in part (C),
since it is rare indeed to find the exact subject matter
sought, the question is the degree of difference over
the prior art which would result in a finding of
patentability. Equivalents are usually particular
examples where this degree of difference is not suffi-
cient to predicate patentability on the ground that
the improvement, although somewhat different
than the prior art, is not inventively different. In
each of Austria, Australia, Denmark, Germany,
Japan, the Netherlands, Norway, Sweden and the

U.S.A., analogous arts are considered and a doc-
trine of equivalents is relied upon. In Canada and
in the United Kingdom, the doctrine of equivalents
is applied cautiously. Canada and the United
Kingdom also rely upon references from analogous
arts but the United Kingdom applies such refer-
ences cautiously. Probably all countries use the
concepts of B and C cautiously. In all countries
the question of whether particular arts are "anal-
ogous" must obviously be decided in each case. In
the U.S.A. the test usually relied upon is to examine
the various arts involved to see whether the prob-
lems in each case are so related that one skilled in
one art would realize that he could find a solution
of his problem in the other art. This involves the
problem that certain areas of two arts may be closely
parallel while other portions of the same arts may
be remote from one another.

6A. The following relate to your presently avail-
able facilities for manually searching documents or
their facsimiles. A. Have you adopted the Inter-
national Patent Classification System? Have you
used it as a guide or pattern in developing your own
system?

Summary: There was some confusion as to
whether "adopting" the International Patent Clas-
sification System meant that the patents would be
physically divided into groups in accordance with
I.P.C. for search purposes or whether another
system would be used for this purpose and the pat-
ents merely marked with the I.P.C. along with
that primarily employed. With this in mind, Aus-
tralia, Denmark, Germany, the Netherlands, Nor-
way, Sweden and the United Kingdom have
indicated that they have "adopted" the Interna-
tional Patent Classification System which is inter-
preted to mean that their patents, as they currently
issue, at least have the I.P.C. number printed
someplace thereon. Austria, Canada, the Inter-
national Patent Institute, Japan and the U.S.A.
have indicated that they have not "adopted" the
I.P.C. System. Denmark, Germany, Norway, the
Netherlands and the United Kingdom has indi-
cated that they have used the I.P.C. System as a
guide in "developing" their own systems. Since
it is a huge task to reorganize an existing, compre-
hensive classification system and in view of various
explanatory statements in the answers to this ques-
tion it appears that the existing classification sys-
tems of the above countries are being retained but
any necessary future changes or revisions will be

made in a manner conforming as nearly as possible
with the I.P.C. System. Austria, Canada, Japan,
and the U.S.A. have not used the I.P.C. System to
any extent.

6B. Briefly describe the classification system used.

Summary: With respect to the type of classifica-
tion system used, Denmark, Germany and Nor-
way employ a system in which the "invention" is the
basis of classification. This need not be restricted
to the exact combination recited in the claims and
judgment may be used in interpreting the claims to
decide where the invention really resides. The
claims may be cast in a form which would be clas-
sified in one class but the examiner or classifier will
look beyond the form of the claims and classify the
application (or patent) in a different class which
he feels more closely provides for what he considers
the actual improvement. Since the Netherlands sys-
tem is based on the German system, they therefore
classify the "invention" in the same manner, but
moreover, other striking, newly disclosed matter is
classified. Australia, Canada, Japan, the United
Kingdom and the U.S.A. each has its own system
adapted to its own peculiar needs as dictated by its
own economic system. The U.K. employs a system
in which the "disclosure" is the basis of classifica-
tion, the invention claimed not necessarily determin-
ing the index of the specification. The Australian
system is briefly described in a pamphlet entitled
"Classification of Subject-Matter for Patents of In-
vention" dated October, 1960. In the U.S.A., the
claimed matter constitutes the basis for classifica-
tion. If the claim is not considered commensurate
with the invention, the examiner may try by various
rejections and examination procedures to get the
applicant to substitute claims which more nearly
conform to the examiner's view of the invention.
The applicant however may not agree with the ex-
aminer as to the exact scope of the invention and
may resist this rejection with various recourses to
appellate tribunals. Canada, the United Kingdom
and the U.S.A. state that they have definitions with
explanatory notes for the various classes and head-
ings or subclasses. It is probable that many coun-
tries employ some definitions since the titles of the
schedules are necessarily brief and concise for rapid
visual selection. The type of subject matter selected
for classification obviously determines its value to
the searcher and is influenced by the type of claims
or inventions to be classified, the philosophy of the
classifier, the statutes governing patents and the

227

decisions of the courts having jurisdiction over the
patents issued. Since the different classes in the
U.S.A. system are diverse in nature and format some
experience has been gained in the use of the various
possible schemes. It has been the experience of ex-
aminers in the U.S. that classes which use concepts
based upon obvious characteristics of a particular
art permit extremely simple and quick classification
but are relatively useless for finding the type of art
necessary to examine applications where there is
novelty present. No substitute has been found for
careful evaluation of the problems presented in the
claims and selection of unique features which dis-
tinguish the invention from all others and which
features will generically provide a search field for
species thereof as they appear in new applications.
Exceptions are found in most classes in the form of
"art groups" in which various problems in a par-
ticular industry are accompanied by other problems
which seem to have no existence apart from the
particular industry and defy analysis on their own
merits. These groups usually constitute the excep-
tion rather than the rule in our recent classes. Can-
ada on one hand preferred to use classes based on
either scientific and engineering concepts or func-
tional or utility concepts related to the intended use
of the invention. The Netherlands, on the other
hand, stresses classification based on function more
than Germany but less than the U.S.A. The In-
ternational Patent Institute uses the patent clas-
sification system of The Netherlands. The U.S.A.
and probably most other countries fall between these
extremes and use classes of both types. The United
Kingdom indicates that they have classes which have
been subdivided into "Headings" which are then
further subdivided as necessary. There is stated
to be a "tendency towards general headings rather
than headings restricted to a particular field." The
final subdivisions of the heading have "press-marks"
on "classification marks" allotted to them and cor-
respond either to search files or to clips (indicated by
colour, position or both) on search files. Each
specification together with its drawings and a printed
abridgement is mounted on a card along the top
of which are the clips indicating the appropriate
"press marks." Normally, the entire drawer file
does not need to be perused but only those cards
having the appropriate clips. This serves, in part,
as a substitute for physically separating the cards
into separate bundles and in many cases eliminates
the need for cross referencing from one such bundle
to another where plural concepts appear. It also,

moreover, has the advantage of enabling combina-
tion clip searching to be employed. In the U.S.A. :
(a) Technology is divided into five general areas of
knowledge, (b) Each area is further divided into
an hierarchically arranged number of classes or main
subjects fields. At present some 309 are in use, each
with subject matter definition limiting its cope and
having notes giving related subjects fields. (c)
Each class is further divided into an ordered ar-
rangement of subclasses, superiority of subject mat-
ter being indicated by the order of the subject in the
plan or scheme, the first appearing at the top of
the schedule being superior to the last appearing.
At present the system includes some 59,000 such sub-
divisions, each with definitions and notes to in-
dicate scope and related or excluded subject mat-
ter, (d) The system looks to identify relationships
of concepts and constituent parts on as near a
proximal function basis as possible, (e) The title
of the subclass indicates the subject matter included
and patents are thus not excluded therefrom because
of some additional function, structure or relation
unless the title specifically uses words limiting the
scope thereof.

6C. How is the system used? (e.g. in determining
placement of patents and cross referencing) .

Summary : It appears that many countries answer-
ing this question collect patents of each subclass (or
subdivision) number together into a group or bundle
which may be inspected as a group to search con-
cepts provided by the particular subclass. Cross
references i.e., additional copies, are placed in other
subclass groups where they would be of interest to
one searching the other subclass. Germany alone
noted that they used "only one classification symbol
with no reference". However this refers only to the
use of such symbols. In the search files every speci-
fication is cross-referenced as required. Sweden
follows Germany in this respect. Norway employs
cross references and prints them on the patents. The
examiners decide which cross references should be
printed. The Netherlands employs cross references
using some complete and some partial copies. Den-
mark employs cross-referencing to a limited extent.
In Japan, the classifier places a rough classification
showing class and subclass on an application and
after examination the examiner gives a subdivision
mark, noting desired cross-references, if necessary.
The classification examiner finally checks the clas-
sification mark. In Australia the system is used to
classify the pending applications for examination

228

and for collecting patents, abstracts and the like into
groups having the same class number for searching.
Each disclosure may have any number of "class
marks" or press marks as both claims and disclosures
are classified. In the U. K., the whole of the dis-
closure is considered for index, "cross-referencing"
(in the U.S.A. sense) being used to a considerable
extent and being limited solely by the defined rela-
tionships betwen headings and by the examiners,
who are at liberty to decide that repetition in any
particular heading would not be fruitful. Com-
plete copies of the specification (and abridgement)
are almost always placed in each file in which the
specification has been indexed. As in the U.S.A.,
lists containing the serial numbers of all patents clas-
sified in each subdivision are tabulated by machine
and are on sale to the public. In the U.S.A. lists
containing the numbers of all patents classified in
each subclass are tabulated by machine and may be
purchased for use in searching patents arranged
in numerical order in various libraries throughout
the U.S.A. With respect to the placement of patent
copies in the U.S. Classification system: (a) The
claims of the patent precisely define the difference
or advance over what was previously known. Based
on the patentee's definition of the invention, the
placement of the advance in the hierarchical scheme
is determined, (b) Where claims to different classes
of inventions are present in the same case precedence
is given in the following order :

( 1 ) Process of using Article

( 2 ) Article used or Manufactured

( 3 ) Process of Making Article

( 4 ) Machine for Making Article*

(5 ) Composition or intermediate subcombina-

tion
*The Machine for Making the Article (item
4 ) is itself an article and would move up to item
(2) if it were the item made by item 3 or by
another machine (item 4) ).
(c) Claims to the type of subject matter which
appears first in the above paragraph ( b ) are said to
control "original" classification and will cause one
copy of the patent to be encoded to an appropriate
class and subclass for that item. Other copies
("cross-references") will also be encoded, (d)
Cross-referencing falls into two procedural cate-
gories, i.e. — mandatory and optional. Mandatory
cross references result from claimed subject matter
for which a subordinate subclass has been estab-
lished. Optional cross references result from un-
claimed but disclosed subjects matter for which a

locus has been provided in any class, (e) Within
the selected class as determined by (b) and (c)
above, the encoder starts at the top of the class
scheme or schedule of subclasses and rejects each
of the coordinate main line concept designations
until the subclass as titled and defined includes the
subject matter of the patent claim. Each subclass
has a definition which recites a unique concept and
distinguishes the concept from that of other
subclasses.

6D. How is the classification system controlled
and maintained on a current basis?

Summary: With respect to the problem of keep-
ing the various classification systems current in order
to provide for new and different types of inventions;
Canada and the United Kingdom make revisions
and additions either because the classification divi-
sion recognizes the need or because the examiners
suggest the change due to search difficulties. In
Australia, new classes are opened up infrequently
and the tendency is to try to force the application
and patents into the existing classification frame-
work wherever possible. Germany makes revisions
as necessary and all revisions are made "in collabo-
ration with the Working Party for Classification of
the Committee of Experts on Patents of the Council
of Europe" and conform generally to the Interna-
tional Patent Classification System so that these sys-
tems should tend to approach the I.P.C. system
after a period of time. The Netherlands makes
revisions if necessary with respect to search. If an
I.P.C. draft of sub-division is available they strongly
urge a revision according to that draft. If such a
draft is not available they rarely discuss matters in
the Working Party, but follow the general trend of
the practice of the Working Party. Denmark,
Norway and Sweden keep their systems up to date
by adopting the current changes and revisions made
by the German Patent Office, whether they have an
immediate need therefore or not. Japan's reply
indicates that major revision of their entire classi-
fication system is attempted occasionally by adding
new classes, subclasses and subdivisions to the pres-
ent system. On the other hand, the U.S.A. replaces
outmoded classes which are still active by more
modern classes on a piecemeal basis. Several classes
may be abolished or combined and re-established
as a new class or, in many cases, the class is merely
re-arranged to make the material classified therein
more available. During revisions, patents which
belong in other classes due to errors or to interven-

229

ing changes in other classes are transferred to the
other classes. The classifiers are grouped into five
divisions each having specialized jurisdiction of a
related group of classes. Different classes are re-
vised simultaneously, each revision constituting a
project having its own leader and a number of
classifiers dependent upon the size of the project.
The Japanese Patent Office has indicated that they
plan to make a reference book to correlate their
classification with foreign systems. It is presumed
that this eould also be in the form of search notes
for each subclass in any system which would advise
a searcher to continue the search in certain stated
subclasses in each of various foreign classification
systems. The U.S.A. and other patent offices (e.g.
The Netherlands) have attempted to solve this
problem by incorporating foreign patents into their
own systems and collecting them in drawers or bun-
dles contiguous to their own patents and cross-
references of the same subclass. The foreign clas-
sification is of some help in this operation where
there is a language problem. The Netherlands has
indicated that they use "partial copies" extensively
as stated above. In the U.S.A. such partial copies
are referred to as "digests of patents" and are usually
reserved for small but important portions of rather
large patents in order to conserve space and time
in locating the pertinent matter in the patent. The
term "digest" is unfortunate since this term is also
used to designate unofficial collections made by ex-
aminers for use by themselves and other examiners.
Persons outside the Patent Office are usually allowed
to search the latter "digests" although they exist in
the examining division only. There is no conten-
tion that such digests provide a complete search.
Sweden indicates that examiners subdivide patent
material for their own use. In the U.S.A., the
Patent Office classification is maintained by about
fifty professional members of the classification group
who put their work into the system each year revis-
ing, updating, and creating classifications. This
does not include the encoding of some 50,000 new
patents which issue each year, nor other technical
literature and foreign patents which are classified
into the existing system by other employees in the
Office.

6E. Which of the following are included in your
plans for future activity in this area? I. Improv-
ing or reworking existing classification schedules.
2. Adopting new theories or differently based clas-
sification schemes.

230

Summary: As to whether the various countries
intend to maintain their present systems or adopt
new theories or differently based classification
schemes, Australia, Austria, Canada and Japan
state that they will maintain their present system,
where acceptable, but look for new classification
ideas at all times and revise old classification sched-
ules in accordance with the plan most suitable to
the particular subject matter involved. Denmark,
Germany, the Netherlands, Norway and Sweden
are changing over to the I.P.C. system as revisions
are made.

6F. 1 . Are foreign patents filed in their national
classification scheme? 2. If the answer is "no",
how are foreign patents classified in your existing
system? 3. What foreign collections of patents do
you have?

Summary: As to the manner in which foreign
patents are filed in the various Patent Offices: —
Austria, Germany, the Netherlands and the U.S.A.
file foreign patents in the same classification system
as their own patents and presumably search them
together. Japan has the patents of 39 foreign coun-
tries. Such patents are arranged in numerical
order, except those of the U.S., United Kingdom
and Germany, which are classified according to
their own classification. Canada files the U.S. pat-
ents in accordance with the U.S. classification
printed thereon at issue and has the patents of
other countries arranged in numerical order. Nor-
way files British and French patents in The Neth-
erlands' system, U.S. and Austrian patents in their
own national classification scheme and most other
patents under the German system. Each country's
patents are arranged in separate cases marked by
different colours. Denmark files patents in their
national classification systems except that the Brit-
ish and French patents are filed according to the
Netherland's system, similarly to Norway. Aus-
tralia does not search foreign patents and therefore !
has no need to keep them in any classified order.
Sweden arranges the U.S. patents in accordance
with the U.S. system and all others are assimilated
into the Swedish system. The United Kingdom
absorbed older U.S. patents (prior to about 1940)
into their national system, one copy being encoded
by the examiner on a best utilized basis, and main-
tains in their Library numerically arranged collec-
tions of patents of other countries (in the case of
recent German specifications, the Library has them
arranged in their own national classification sys-

tem) since, as stated above, the latter patents are
not usually searched. Regarding the number of
countries whose patents are on file at the various
patent offices, no particular pattern is apparent ex-
cept that ( 1 ) most patent offices appear to have
foreign patents from Belgium, France, Germany,
Switzerland, the United Kingdom and the U.S.A.
( 2 ) Many countries have on file patents from neigh-
boring countries and countries having similar lan-
guages or economies.

6G. What aids to manual searching are utilized?

Summary: Aids to manual searching. This area
is construed to include all devices and expedients
which would aid in selecting actual patents, photo-
graphic or printed reproductions thereof or ab-
stracts thereof, each of which would have to be
visually inspected for selection or rejection. .Def-
initions may be considered an aid to searching and
these have been mentioned above. An index may
also be of value to locate concepts but semantic
problems limit its use. Canada, Germany, the
Netherlands, the U.S.A. have such indices and the
International Patent Classification System has a list
of "catch-words". The problem is that the same
word has different meanings to different people; not
only to people in different countries but people in
different industries or in the same industry at dif-
ferent times, even in the same country. Such an
index could provide a central place to define each
word indexed as used in each of its occurrences but
this apparently has not been done. It would also be
useful to use ample cross index or reference items
for synonyms. Japan stated that they used abstracts
on search cards. The U.K. uses abridgements as
well as the original classifications on their search
cards and also uses annotations ( lists, flow-diagrams,
etc.) as well as the filing and clipping system re-
ferred to above under 6B. Australia employs ( 1 )
index cards for each class subdivision having
thereon the application numbers of all disclosures
having the subject matter provided for by the sub-
division, (2) classified sets of patents or abstracts (3)
classified records of pending, unpublished applica-
tions, and (4) search notes privately maintained by
the examiners.

Cards with abstracted information and possibly
with a small reproduction of a patent thereon may
be searched by various means. A common way is
to assign a symbolic meaning to each of numerous
holes at the edges of the card. Each card would
then have the edges removed except for the portion

having a hole or plural holes corresponding to an
encoded disclosure in the reference. The cards
may then be selected by passing a long rod thru the
aligned holes corresponding to a concept sought, in
adjacent cards. The rod would then remove only
the desired cards from the pack. "Peek-a-boo"
cards could be similarly arranged but selected in
response to passage of a light beam thru a specified
portion of the card corresponding to the concept
desired. The International Patent Institute has a
punched card digest for luminescent material and
is completing another for chemical treatment of
textiles and a Batten card digest for laminated ma-
terials. Germany uses manually selectable
punched cards for patents relating to breweries,
pumps, steroids, locks, printed circuits, aircraft en-
gines, building construction parts. In the Nether-
lands, where multiple classification is thought nec-
essary within restricted fields only, it is used without
the need of more than one copy by means of a spe-
cial arrangement of the specifications. This is
done by fixing the order of any possible combina-
tion of search features of the system and arranging
the specifications accordingly. The specifications
on electrical control and on automatic telephones
are made accessible by use of multi-column scan
lists. Those lists have one column for the specifi-
cation number and one for each of the search fea-
tures. There is also a special card system on azo
dyes. The United Kingdom plans to use Batten
cards extensively and presently has collections of
luminescent materials and packaging and has in
preparation digests of recording, textiles and cath-
ode ray tubes. An additional digest on steroids is
being considered. A modified Brisch-Vistem de-
vice will be used for these searches. The United
Kingdom also plans to use "Findex" punched cards
in cooperation with a micro print reading device.
Australia plans to investigate the Batten (inverted
file) manual searching system. It would seem that
punched cards of this type could be transcribed
mechanically at a later date if sorting machines or
computers eventually were used. On a trial basis
in the U.S.A., microfilm cards have been substi-
tuted for the patent copy and an optical reader is
used in searching. In some examining divisions an
abstract or digest of the patent is written out on a
card and then cards are filed (unofficially) accord-
ing to some plan and then used as a basis for search.
In at least one area a "finder list" is developed for
each patent, appropriate places are checked for a
developed scheme of rather frequently occurring

645803 O-63-ll

231

types. The lists are then searched to identify pat-
ents having the type feature desired.

7 . The following relate to mechanized searching.
A. Do your search facilities include a mechanized
search system? B. Give a brief description of the
mechanized system employed, the equipment used
and the area(s) of technology to which it is ap-
plied. C. What are your present activities in and
future plans for conducting research in the mech-
anization of information storage and retrieval?

Summary: With relation to mechanical search-
ing systems it is noted that some of the devices con-
sidered under question 6G could be considered
"mechanical searching devices" from some points
of view. Remaining for consideration are the more
or less "automatic" devices for selecting cards
manually placed into the machine. The Inter-
national Patent Institute is preparing a punched
card search file from the multi-column scan list on
electrical control in the Netherlands Patent Office
which is to be searched on a multi-column sorter.
Japan is investigating the use of mechanized search-
ing systems and plans some experiments in mechan-
ical searching. The United Kingdom has cards
for computer searching on "packaging" and on
"coating with metals". Denmark, Norway and
Sweden are searching U.S. cards on Steroid com-
pounds and when available may use U.S. cards for
"Organic phosphorus compounds" (U.S. class 260,
subclass 46 1 ) . Norway employs a "Norsk Regne-
central" — Bull D-3 Electronic Computer while
Sweden uses IBM type 85 and type 101 machines.
No present plans have been reported to extend this
type of searching. The Netherlands reports that
they have mechanical searching plans under study
but are not presently using this type of equipment.
Australia has no present research activity. With
respect to work on mechanized searching in the
U.S.A., three mechanized searching systems are in
current use:

1. Steroids have been searched on an IBM-101
since August, 1957. Direct coding is used, each
punch position of a standard IBM card being re-
served for representation of a particular feature
which is superimposed on an assumed, fixed steroid
nucleus. For example a punch in one particular lo-
cation indicates a double bond at position 5 . The file
includes U.S. patents and a substantial coverage of
periodical literature.

2. Organic phosphorous compounds are searched
by a system called RAMP which uses the IBM

RAMAC-305. Partial use of this system began in
January 1961 and full operational use has been in
effect since October 1 of this year. A punched card
version of this system, called CAMP, is also used,
employing the IBM-101. The file, which includes
about 2000 patents, is in the inverted file form. En-
tries are based on functional groups and the most
specific level of the file lists documents describing
compounds containing all of the groups asked for,
a list of document numbers is printed out.

3. Transistor circuits are searched by a peek-a-
boo system that now contains about 900 patents in
the file. Another 1100 patents are being analyzed
and coded for addition to the file. The topology
of the circuits and some of their general character-
istics are recorded. Access to copies of references
found by the search is by means of a Lodestar Micro-
film Reader. Consideration is being given to the
use of the Microcite Reader developed by the Na-
tional Bureau of Standards.

A number of experiments are in progress dealing
with a variety of problems and encompassing a num-
ber of subject matter fields.

( 1 ) In the field of the mechanical arts, a method
of recording and searching hydraulic circuits is be-
ing developed. Symbology for the mechanical ele-
ments and fluid flow-paths are being worked out in
such a way as to permit tracing through any of
the operational sequences present.

( 2 ) A hand sorted punched-card system for re-
cording information about artificial respirators is
being developed.

( 3 ) A system known as PACIR, which will store
and retrieve patents on pesticides, is now in the de-
velopment stage. Two versions will be available:
one is for use with the IBM-RAMAC 305 ; the other
being a punched card system for use with the IBM-
101.

(4) On the HAYSTAQ project, a computer
system for chemical searching, an intensive testing
and evaluation of the topological structure search
routine has been started. This is designed for han-
dling any compound of known structure and per-
mits considerable flexibility in framing questions.
This is now programmed for SEAC at the Bureau
of Standards, but consideration is being given to
reprogramming for the 7090. Work was begun
this summer (1961) on means for enabling com-
puters to do most of the encoding of information
extracted from the documents by trained chemists.

( 5 ) An experiment has recently been started in
which the full natural text of 100 transistor patents

232

will be made available for processing by a general
purpose computer. A variety of approaches to
automatic indexing and full text searching will be
used and several linguistic techniques will be
employed.

(6) Notation systems are being developed to
enable a computer to store and retrieve complete
electronic and electrical schematic diagrams. Lin-
ear, lattice and matrix notations are being investi-
gated.

( 7 ) A number of other projects, not immediately
applicable to searching systems, have been, or are,
in progress, (a) A system of symbolizing chem-
ical reactions and processes, which emphasizes the
sites of reaction, was developed and a report is
available, (b) A new, comprehensive enumera-
tion and linear notation system for chemical com-
pounds has been worked out and will be described
in a publication which is in preparation. This
system is' such that every compound has a unique
cipher and every cipher can be translated unam-
biguously to a unique structure. The regularity of
the method has permitted development of an al-
gorithm for converting a linear cipher to a cathode

ray tube display of the corresponding structure.

(c) Some work has been done on the application of
document storage and search techniques to search-
ing for legal decisions on certain aspects of the pat-
ent law, particularly as related to design patents.

(d) Work of a more theoretical nature on linguistic
problems, particularly various relationships indi-
cated between elements of information in a text,
and the study of prepositionals in the English lan-
guage, is not at the moment active. This work
was directed at the ultimate establishment of an
unambiguous metalanguage for storage of docu-
mentary information, (e) Some support has also
been given to the investigation of formal linguistics
and the processing of pictorial information which is
being conducted at the Bureau of Standards.

The achievement of a fully mechanized infor-
mation retrieval system for Patent Office use cannot
be realized without solution of many basic problems
of a theoretical nature, as well as system and hard-
ware advances. It is anticipated that there will be
considerably expanded effort in these directions
while continuing to make use of presently available
techniques.

233

APPENDIX C

ATTENDANTS

at the

INTERNATIONAL PATENT OFFICE WORKSHOP

ON
INFORMATION RETRIEVAL

AUSTRALIA

Mr. Cecil N.Jeff ery
Supervising Examiner
Classification Division
Patent Office
Canberra, A. C. T.
Commonwealth of Australia

CANADA

Mr. J. W. T. Michel
Commissioner of Patents
Patent and Copyright Office
Ottawa, Canada

Mr. F. W. Simons
Assistant Commissioner of Patents
Patent and Copyright Office
Ottawa, Canada

FEDERAL REPUBLIC OF GERMANY

Dipl. Ing. Werner Rubach

Direktor im Deutsches Patentamt

Zweibruckenstr, 12

Munich 2

Federal Republic of Germany

JAPAN

Mr. Kotaro Otani

Chief Examiner

Second Examination Division

Patent Office

1 , Sannen-Cho, Chiyoda-Ku

Tokyo, Japan

THE NETHERLANDS

Dr. C. J. deHaan

President

Octrooiraad en Merkenbureau

Willem Witsenplein 6

The Hague, Netherlands

Mr. Albert J. Bulder

Vice-President

Octrooiraad en Merkenbureau

Willem Witsenplein 6

The Hague, Netherlands

Mr. P. van Waasbergen

Vice-Director

Institut International des Brevets

97 Nieuwe Parklaan

The Hague, Netherlands

NORWAY

Mr. Johan Helgeland

Director

Sytret for det industrielle rettsvern

Middelthunsgt. 15 B

Oslo, Norway

Mr. Knut Ueland

Chief Engineer

Styret for det industrielle rettsvern

Middelthunsgt. 15 B

Oslo, Norway

234

PHILIPPINES

Mr. Tiburcio S. Evalle

Director of Patents

Patent Office

Department of Commerce and Industry

Boston Street

Port Area

Manila, Philippines

SWEDEN

Mr. Claes Uggla

Head of Law and International Affairs

Kungl. Patent-och Registreringsverket

Box 5055

Stockholm 5, Sweden

Mr. Tore Oredsson,

Bureau Director

Kungl. Patent-och Registreringsverket

Box 5055

Stockholm 5, Sweden

UNITED KINGDOM

Mr. L. F. W. Knight
Superintending Examiner
The Patent Office
25 Southampton Buildings
London W. C. 2, England

UNITED STATES

Department of Commerce, Patent Office

David L. Ladd,
Commissioner of Patents

Edwin L. Reynolds,

First Assistant Commissioner of Patents

Horace B. Fay, Jr.,

Assistant Commissioner of Patents

Don D. Andrews,

Office of Research and Development

C. D. Angel,
Division 95

Walter Berlowitz,
Division 94

R. Birch,
Scientific Library

W. R. Campbell,
Scientific Library

P. C. Coffin,
Scientific Library

W. S. Cole,
Division 94

E. C. Darsch,
Division 95

P. J. Federico,
Examiner in Chief

Julius Frome,

Office of Research and Development

Charles F. Gareau,

Division 22; Chairman, Working Committee, 125th
Anniversary Committee

Harry Gauss,
Division 92

George A. Gorecki,
Classification Group

Winston Hayward,

Office of Research and Development

E. A. Hurd,

Organization and Methods Division

Irene Jacobs,
Scientific Library

Robert Johnson,
Division 92

Arthur P. Kent,
Division 91

J. R. Kline,

Office of Research and Development

Herbert R. Koller,

Office of Research and Development

Ivan Lady,
Division 91

Jacob Leibowitz,

Office of Research and Development

Harold F. Lindenmeyer,
Scientific Library

235

E. R. Mackert,
Division 94

Paul T. O'Day,

Office of Research and Development

Kingsley C. Peck,
Division 93

Walter A. Pennino,

Office of Information Services

Harold Pfeffer,

Office of Research and Development

Jonathan Plaut,
Division 62

Samuel B. Pritchard,
Division 92

W. F. Purdy,
Division 93

Joel Reznek,
Division 95

Paul A. Robey, Jr.,
Division 92

Howard B. Rockman,
Division 12

George C. Roeming,
Law Examiner

I. J. Rotkin,
Division 93

Arnold Ruegg, Jr.,
Division 94

Gladys Sampson,
Scientific Library

Charles G. Smith,

Office of Research and Development

Richard A. Spencer, Jr.,

Office of Research and Development

E. N. Spithas,

Office of Research and Development

Donald P. Stein,

Office of Research and Development

Joseph Stitelman,

Office of Research and Development

Peter Urbach,

Organization and Methods Division

H. S. Vincent,
Division 91

D. A. Waite,
Division 12

M. O. Wolk,
Division 43

Charles Yates,
Division 92

Eugene E. Young,
Division 92

Other Government Agencies, Industry and Universities

Dr. S. N. Alexander, Chief
Data Processing Systems Division
National Bureau of Standards
Washington 25, D.C.

Dr. Roy C. Amara, Manager
Systems Engineering Department
Stanford Research Institute
Menlo Park, California

Dr. Julian Bigelow
Institute for Advanced Study
Princeton, New Jersey

Mr. Charles P. Bourne
Research Engineer
Stanford Research Institute
Menlo Park, California

Dr. Harold F. Bright

Patent, Trademark, and Copyright Foundation

The George Washington University

2 107 H Street, N.W.

Washington, D.C.

Mr. Julius N. Cahn

Staff Director

Subcommittee on Reorganization

Senate Committee on Government Operations

Old Senate Office Building

Washington 25, D.C.

Dr. Watson Davis
Director, Science Service
1719 N Street N.W.
Washington 6, D.C.

236

Mr. R. A. Fairthorne
Visiting Research Professor
Western Reserve University
Cleveland 6, Ohio

Mr. M. C. Freudenberg
Air Force Patents Division
Room 4322
Munitions Building
Washington 25, D.C.

Mr. Ezra Glaser, Section Chief
Data Processing Systems Division
National Bureau of Standards
Washington 25, D.C.

Mr. John C. Green
Director, Office of Technical Services
Department of Commerce
Washington 25, D.C.

Mr. L. James Harris

Patent, Trademark, and Copyright Foundation
The George Washington University
Washington 6, D.C.

Dr. L. B. Heilprin
Council on Library Resources, Inc.
1025 Connecticut Avenue
Washington 6, D.C.

Dr. Karl Heumann
National Academy of Sciences
2101 Constitution Avenue, N.W.
Washington 25, D.C.

Mr. J. Jancin, Jr.

Washington Patent Operations Manager

International Business Machines Corporation

425 13th Street, N.W.

Washington 4, D.C.

Mr. Martin Kalikow, Chairman
Subcommittee on Mechanized Searching
American Patent Law Association
General Electric Company
570 Lexington Avenue
New York 22, New York

Dr. Harold Kantner
Armour Research Foundation
Illinois Institute of Technology
Chicago 16, Illinois

Mr. Russell Kirsch
Data Processing Systems Division
National Bureau of Standards
Washington 25, D.C.

Mr. Kay I. Kitagawa
National Science Foundation
Washington 25, D.C.

Mr. Reed C. Lawlor
Suite 922

Park Central Building
412 West Sixth Street
Los Angeles 14, California

Mr. Joseph M. Lightman
Department of State
Washington 25, D.C.

Miss Judith T. MacMillan
Assistant Chief, Washington Branch
Institute for Advancement of Medical Communi-
cation
1028 Connecticut Avenue, N.W.
Washington 6, D.C.

Mrs. Ethel C. Marden
Data Processing Systems Division
National Bureau of Standards
Washington 25, D.C.

Mr. Simon M. Newman

1411 Hopkins Street, N.W.
Washington 6, D.C.

Mr. Eugene Pronko
National Science Foundation
Washington 25, D.C.

Mr. John Robb

Electronic Data Processing Division
Radio Corporation of America
Camden 8, New Jersey

Dr.J.H.Shera

Dean

School of Library Science

Western Reserve University

Cleveland 6, Ohio

Mr. John Sherrod, Jr.

Chief, Science and Technology Division

Library of Congress

Washington 25, D.C.

237

Dr. Irving H. Siegel, Chief

Military Economics and Costing Division

Research Analysis Corporation

6935 Arlington Road

Bethesda 14, Maryland

Mr. Richard A. Spencer, Jr.
1 12 Newlands Street
Chevy Chase, Maryland

Mr. Allen J. Sprow
National Science Foundation
Washington 25, D.C.

Miss Mary E. Stevens
Data Processing Systems Division
National Bureau of Standards
Washington 25, D.C.

Mrs. Rowena W. Swanson
Air Force Office of Scientific Research
Room 2406, Tempo D
Washington 25, D.C.

Dr. Fred Tate

Assistant Director

The Chemical Abstracts Service

The Ohio State University

Columbus 10, Ohio

Robert C. Watson
Former Commissioner
U.S. Patent Office
Washington, D.C.

Dr. C.J. Wessel, Director
Prevention of Deterioration Center
National Academy of Sciences
National Research Council
2101 Constitution Avenue
Washington 25, D.C.

Mr. R. E. Wiley

Advanced Systems Development Division
International Business Machines Corp.
Yorktown Heights, New York

Dr. Harold Wooster
Director of Mathematical Sciences
Air Force Office of Scientific Research
Room 2406, Tempo D
Washington 25, D.C.

Mrs. Helen L. Brownson
National Science Foundation
Washington 25, D.C.

Mr. Ben Chromy
Post Office Box 905
Los Gatos, California

Mr. A. Katz

RCA Advanced Military Systems

Camden 8, New Jersey

Mrs. Louise Sissman
Department of State
Washington 25, D.C.

Mrs. Ruth E.Suse

National Bureau of Standards

Washington 25, D.C.

238

APPENDIX D

COMMITTEES

Policy Committee

G. A. Gorecki, Chairman
E. Glaser
I. Marcus
H. Pfeffer

Arrangements Committee
H. Pfeffer, Chairman
H. W. Hayward

Mrs. M. G. Hepler
Mrs. A. T. Jaramio
H. R. Roller
E. F. Luckett
P. T. O'Day
A. Ruegg, Jr.
E. N. Spithas
D. P. Stein
R. A. Wahl

239

U.S. GOVERNMENT PRINTING OFFICE 1963 O — 645803

PENN STATE UNIVERSITY LIBRARIES

AQQQt^QIQS^b