MACHINE TRANSLATION

• Document Type: CREST
• Collection: General CIA Records
• Document Number (FOIA) /ESDN (CREST): CIA-RDP80B01139A000200040009-9
• Release Decision: RIPPUB
• Original Classification: K
• Document Page Count: 12
• Document Creation Date: December 15, 2016
• Document Release Date: April 7, 2004
• Sequence Number: 9
• Publication Date: May 12, 1960
• Content Type: MF

Approvedor Release 2004/05/05: CIA-RDP80BQ139A000200040009-9
COD1B?D-56
12 May 1960
U NITED S T A T E S I N T E 1, L I C E N C E B O A R L
Pursuant to the Twenty-second Meeting of the Committee on I"u-
mentationo attached for your information is a copy of n5r statement
to be made on mechanical translation before the House Committee on
Science and Astronautics,
Paul A. Borel
Chairman
Approved For Release 2004/05/05 : CIA-RDP80B01139A000200040009-9
 Approved For Re se 2004/05/05: CIA-RDP80B01139A 0W200040009-9
Statement of Paul A. Borel, Assistant Director
for Central Reference, Central Intelligence
Agency, and Chairman of Documentation Committee
of the United States Intelligence Board, before
the House Committee on Science and Astronautics,
May 13, 1960.
Mr. Chairman and Members of the Committee, the Central Intelligence
Agency is pleased to respond to your invitation to outline its views on
machine translation.
Our interest in MT dates from 1951, when some of our scientists dis-
cussed the possibility of developing an automatic indexing and trans-
lating machine with Dr. James Perry, then with the M.I.T. Center for
International Studies and now Director of the Center for Documentation
and Communications Research, Western Reserve University. After some pre-
liminary work, Dr. Perry and CIA representatives in June 1952 attended a
meeting at M.I.T. of linguists, logicians and mathematicians on the subject
of machine translation. The principal result of that meeting, which was
promoted by Dr. Bar Hillel and supported by the Rockefeller Foundation,
was the further stimulation of interest and the realization of possibility
in the minds of some of the linguists present.
In the next two years or so, CIA reviewed various proposals, includ-
ing proposals from M.I.T., the Battelle Memorial Institute, and Georgetown
University. Some of these were considered jointly with elements of the
Department of Defense.
Our position during that period was that the development of a machine
translation capability was highly desirable, and hence that we should
Approved For Release 2004/05/05 : CIA-RDP80B01139A000200040009-9
 Approved For Re1se 2004/05/05: CIA-RDP80BO1139A 2200040009-9
support an Iva program. We recognized, however, that such a program had im-
plications which transcended the interests of CIA and those of the intelli-
gence community. We therefore considered it preferable that an organization
with broader responsibilities than our own be prevailed upon to take the
initiative to push a comprehensive MT program. We identified our immediate
need as a usable product, i.e., one which might well be far short of a
perfect translation but nevertheless highly useful. In return for an early
MT capability to produce a usable product, we were willing to leave the
achievement of superior results to a longer range program.
This pragmatic approach was our aim and purpose in 1951. It remains
our aim and purpose today.
In early 1955, CIA approached the National Science Foundation, and
concurrently ascertained the degree of interest in the Department of
Defense. These overtures were directly related to one of a succession
of proposals by Professor Leon E. Dostert of Georgetown University.
1W
Defense representatives were "all in substantial agreement that, while
the Department of Defense does not find it possible to authorize any
funds for this project, we will be very much interested in any such
device once its feasibility has been firmly established". Negotiations
with the National Science Foundation culminated, in early 1956, in an
exchange of correspondence between Dr. Alan T. Waterman, Director,
National Science Foundation, and Mr. Allen W. Dulles, Director of
Central Intelligence. The National Science Foundation agreed r"to
administer any part of a program of research in machine translation
which is agreed by all concerned to be desirable". CIA recognized
the need for careful planning and coordination "to insure maximum
Approved For Release 2004/05/05 : CIA-RDP80B01139A000200040009-9
 Approved For RLase 2004/05/05 : CIA-RDP80B011390 0200040009-9
-3-
ANO progress toward our immediate goal of a machine capability to translate the
Russian technical literature".
Two short excerpts from Mr. Dulles' letter of February 29, 1956, are
useful to point out some of the broad considerations:
"I should like to reaffirm the deep interest which we
in the intelligence field have in the possibility of
translation of Russian language materials, particularly
in scientific fields, into English by machine. In
addition, many of us feel that the degree of human
understanding that could be accomplished if language
barriers could be lowered without sacrificing linguistic
integrity might well be a major step toward peace."
"It is our opinion that much is to be gained by the early
development of a machine capability for translation. The
National Security can be well served if we have available
the scientific and technical literature of the USSR in
English for detailed analysis as early after publication
as possible. I am assured by leaders in electronic research
that technological problems yet unsolved need not stand in
the way of the rapid development of a machine once the
linguistic research had been started."
In the period between May 1956 and the present, the U.S. Government has
provided financial and logistical support for the Georgetown project total-
Department of Defense
(estimated value of
computer time provided
by Air Force and Army)
CIA grants (direct or
reimbursement to NSF)
CIA assistance in text
preparation
Considering the inherent difficulties of the undertaking, we have been
very pleased with the substantial progress made by Georgetown University.
25X1
Approved For Release 2004/05/05 : CIA-RDP80B01139A000200040009-9
 Approved For R ase 2004/05/05 : CIA-RDP80B01139AW200040009-9
Eff~.:;s to date have been experimental, with Russian organic chemistry
texts processed totaling about 500,000 words, and those in
physics some 200,000. It would be extremely valuable to apply the lessons
learned to the automatic translation of several million words of texts
covering various diciplines of particular interest, such as, organic
chemistry, geophysics (astronomy, meteorology and celestial mechanics),
physical chemistry, high energy physics and solid state physics. And
in this connection we are currently studying a proposal by Georgetown
University to conduct a large scale operational feasibility test during
the next fiscal year. As a further indication of the development in the
state of the art, I should also mention that only last week we received a
proposal by the International Business Machine; Corporation for the establish-
ment of an automatic translation facility. The Committee may be interested
in the exchange of correspondence between Dr. E. R. Fiore, Director of
Research, IBM, and Mr. Dulles, which I will be pleased to make available.
This proposal too is under study.
It is fair to ask whether developments since our initial interest in
MT have called for a change in original objective. In the main, the answer
is no. In recent years the volume of available Russian scientific and
technological literature has greatly increased. We estimate that the
This increase
has been accompanied by increased efforts by the Government to translate
the most useful part of this production. And the performance, 95% of
which is by the Government or under Government contract, is impressive.
About 53 million words of Russian scientific literature are now being
translated annually (of which CIA accounts for over 9 million).
Approved For Release 2004/05/05 : CIA-RDP80B01139A000200040009-9
 Approved For Re Vase 2004/05/05: CIARDP80B01139AQW200040009-9
translate on an average
Why then the need for MT? The reasons can, I think, be simply stated:
(1) The volume of publication will continue to increase,
and at a rate in excess of our ability to procure
competent translations.
(2) The quality of translation work done through contract
arrangements is not uniformly excellent. Whatever the
level of accomplishment in MT at any given time, the
output is uniform. In short, MT holds out the promise
of a uniformly more accurate product.
(3) MT also promises greater speedo We now give priority
to categories and languages of greatest interest. Non-
priority items are invariably slow in reaching the
reader. Perhaps the translator with the particular
skill in a language, or in a discipline, cannot
immediately take on the task. In any case he cannot
And
these rates will increase. Even if post-editing were
required, the man-machine system would appreciably
out-produce the human translator working alone.
() With MT, more translations would be available. This in-
creased availability of translations would itself generate
new and more widespread demands for them. We now strive
to pass over only marginal material, but cannot be sure
that we are invariably successful.
Approved For Release 2004/05/05 : CIA-RDP80B01139A000200040009-9
 Approved For Rel se 2004/05/05: CIA-RDP80B01139AO 00040009-9
6-
(,) Greater availability would result in a better informed
corps of scientists in this country. This would result
in. superior evaluations of scientific and economic
developments in the Bloc than is now possible.
(6) The development of a two-way MT capability would make
possible low cost production of American publications
for sale in underdeveloped countries where low cost
Bloc publications now have an almost clear field to
the detriment of U.S. interests.
(7) And finally', the research done and the techniques
developed for accomplishing translation by machine
would contribute materially to the solution of
problems in the broader field of information
storage and retrieval,, and the emerging field of
language data processing.
A word should be said about some of the problem areas. I am not too
concerned about the technical problems. There are many of course. For
example, an all purpose rapid print reader would certainly be essential
to an efficient system. Today we prepare computer input by punching each
Russian word onto IBM cards and then go from card to magnetic tape. Our
25X1
manner. But good work is being done in the technical area and solutions
will be forthcoming. More basic is the problem of organization. Shall
an MT capability once achieved be exploited by each on his own or should
a central facility serve all? If the latter, who shall set it up, who
shall operate it, and under what terms shall Government and private
interests participate?
Approved For Release 2004/05/05 : CIA-RDP80B01139A000200040009-9
 Approved For ReliWse 2004/05/05 : C19-F De80BO1139AOO 00040009-9
Vlar
It is not too early to start thinking about this. A central facility
is, I believe, indicated but not exclusively so. The enormous potential
output of MT greatly exceeds the present and prospective requirements of
any one part of Government or single private organization. Problems of
procuring and selecting materials to be translated, and of disseminating
translations to those needing them, are very considerable and can most
efficiently and economically be solved centrally. Moreover, the use of
a central facility permits the use of equipment exclusively designed to
produce automatic translations. There are, however, requirements for
accomplishing translations under mobile conditions, or, for fully
utilizing general purpose equipment acquired for processing data rather
than lexical material. Hence there is also continuing need for research
to develop multiple language, multiple discipline MT materials and programs
for translation by general purpose computers.
when MT is discussed there is invariably an expressed interest in what
the Soviet Union is doing in this field. I will not dwell on this except
to say that the Soviets have a program which considerably exceeds our own
in scope and size, and that they are doing very good theoretical work,
though restrictions on the availability of computer time has limited
opportunities to apply theory to practice. Two papers, one by Professor
Oettinger 1/ and one by Dr. Harper ?/, provide valuable assessments of
the Soviet effort. The Joint Publication Research Service series Soviet
Developments in Information Processing and Machine Translation, will also
be of interest to the Committee.
l/ Anthony G. Oettinger, "A Survey of Soviet Work on
Automatic Translation", Mechanical Translation,
Vol. 5, No. 3, December 1958, pp. 101-110.
2/ K T: Harpderor,Jeol este K004/05/0 5in Machi 80B0a1nslaA00o0r200040009-9
RAND Corporation Monograph No. P-1896, 4 Feb. 1960, 17 pp.
 Approved For F ase 2004/05/05 : CIA-RDP80B01139400200040009-9
-8-
I will close my remarks with a word about coordination of MT activities
among Government departments. Until recently only a few departments or
agencies had programs. As the field grew CIA took steps to formalize with-
in the intelligence community the informal channels of communications used
by those in charge of MT programs,, whether in the intelligence or the
research components of their respective departments. A group of experts
was formed whose membership is drawn from the Army, Navy, Air Force, State,
NSA, and CIA, with a National Science Foundation representative as associate
member. The function of this interdepartmental group is: to advise the
Committee on Documentation of the United States Intelligence Board with
respect to all machine translation activities, to coordinate'MT activities
within the intelligence community, and to inform its members of new projects
and of the status of existing projects. For over-all coordination in
matters transcending the interests of the intelligence community, CIA
looks to the National Science Foundation.
Approved For Release 2004/05/05 : CIA-RDP80B01139A000200040009-9
 Approved For Fase 2004/05/05 : CIA-RDP80B01139iW0200040009-9
C
0
P
Y
590 Madison Avenue,
New York 22, N.Y.
April 20, 1960
Mrs Allen W. Dulles
Director,
Central Intelligence Agency
Washington 25, D. C.
Last February IBM had the opportunity to present its
research program on automatic translation of languages to Mr.
Amory and other members of the Agency's staff. Following the
presentation a number of suggestions were made by members of
your staff that we should consider seriously the establishment of
an Automatic Translation Facility.
Apart from the research and development program, we
have made studies of the operational requirements, and following
the suggestions we have coordinated them into a proposal, which
we would be very pleased to present to the Agency at your con-
venience.
A large part of our program is supported by the USAF
Air Research and Development Command. We have received official
permission from ARDC to make this presentation, should you wish
to consider it at this time, and to discuss the possibility of consoli-
dating their program with the Agency's, in the interest of initiating
a national Automatic Translation Facility.
S/ E. R. Fiore
E. R. Fiore
Director of Research
ERP: jms
Approved For Release 2004/05/05 : CIA-RDP80B01139A000200040009-9
 k
C
Approved For F ase 2004/05/05 : CIA-RDP80B01139UP0200040009-9
CENTRAL INTELLIGENCE AGENCY
Washington 25o D.C.
0
P
Y
30 April 1960
W
Dr. E. R. Piore
Director of Research
International Business Machines Corporation
590 Madison Avenue
New York 22., N.Y.
Thank you for your letter of 20 April, in which you offer to present
a proposal for the establishment of an Automatic Translation Facility.
We have since 1952 urged the development of a machine translation
capability, particularly for translating scientific Russian to English.
Since 1956, we have provided financial support for the work done in
this field at the School of Languages and Linguistics of Georgetown
University.
We believe that the progress made to date by Georgetown University
and others working on various aspects of this difficult problem is
sufficiently encouraging to justify a large scale operational feasi-
bility test.
It is therefore most timely to explore prospects for consolidating
programs., and we are pleased to learn of Air Force interest in initiating
a national Automatic Translation Facility.
Since it would be helpful to us to review your proposal at an early
date., Mr. Amory- has been in touch with Mr. John Griffith and a meeting
has now been arranged for Maur 5?
Sincerely,
SI Allen W. Dulles
Allen W. Dulles
Director
Approved For Release 2004/05/05 : CIA-RDP80B01139A000200040009-9
 P%W
Approved For Release 2004/05/05 : CIA-RDP80B0 -11 A00020`0040009-9
v
FOR OFFICIAL USE ONLY
CODIB-D-55
12 May 1960
U N I T E D S T A T E S I N T E L L I G E N C E BOARD
COMMITTEE ON DOCUMENTATION
MEMORANDUM FOR; Committee on Documentation
SUBJECT: Sub-Committee on Classification
1. Pursuant to paragraph 7 of the Minutes of the Twenty-first
Meeting, 31 March 1960 (CODIB-M-21, dated 6 April 1960), Mr. Louis C.
Covell, Chief, Document Division, CIA is designated Chairman of the
Classification Sub-Committee and Mrs. Josephine R. Brahm, Staff Assistant
to Chief, Document Division, CIA is designated Vice-Chairman.
2. The Classification Sub-Committee will be concerned primarily
with establishing procedures for revising the Intelligence Subject Code
and for approving any changes, deletions'and additions thereto. In
addition, the Sub-Committee will be. asked by CODIB to study problems
related to the indexing of intelligence documents and submit recommen-
dations for the establishment of uniform procedures.
3. The Sub-Committee Chairman has asked that the names of the
persons who will represent the USIB Agencies on the Sub-Committee be
submitted so that an organizational meeting--can be called at an early
date.
Approved For Rel1/9105 6IRP80B01139A00.0200040009-9