American Journal of Cornputat ional Linguistics 
Microfiche 37 
7 THE FINITE STRING >- 
NEWSLETTER OF THE ASSOCIATION FOR COMPUTATIONAL LINGUISTICS 
A combined alphabetic list of members of the Associ- 
ation and persons mentioned in the 1974 and 1975 
sets of AJCL, both contributors and authors of ab- 
stracted papers, is in preparation. A new scheme 
of topical categories is being prepared by cluster 
analysis of responses from members, and will be 
used as headings on a set of index guide cards which 
will list articles, abstracts and members by topic. 
The whole package will be distributed about a month 
after this issue. 
AMERICAN JOURNAL OF COMPUTATIONAL LINGUISTICS is published by 
the Center for Applied Linguistics for the AssoCiation for 
Computational Linguistics. 
EDITOR: David G. Hays, Professor of Linguistics and af Com- 
puter Science, State University af New York, Buffalo. 
EDITORIAL ASSISTANTI William L. 'Benzon. 
EDITORIAL ADDRESS: Twin Willows, Wanakah, New York 14075. 
MANAGING EDITORZ A. Hood Roberts, Center for Applied Linguistics. 
ASS I STANT : David Hoffman . 
PRODUCTION AND SUBSCRIPTION ADDRESSt 1611 North Kent Street, 
Arlington, Virginia 22209. 
Coyyright 
1976 by the Associaticn for Computational Linguistics 
TABLE OF CONTENTS 
ACL PRESIDENT 1976 Stanley R. Petrick . . . . . . 3 
PERSONAL NOTES . . . * . . . . . . . . . . . *.. 4 
SEMANTICS Georgetown Round Table a . . . , . . . 5 
AAAS PROGRAM EXCERPTS Science and Expectations . . 7 
NEH Application deadlines . . . . . . . . . . . . . 11 
SCIENCE AND GOVERNMENT Ralston speaks at NCC 76 . . 12 
ARCHITECTURE FOR NONNU'MER3:C PROCESS3NG . . . , . . 13 
ASIS PRESIDENT 1976 Melvin S. Day . . . . . ...- 15 
LITERARY AND LINGUISTIC COMPUTING Meeting notes . . 16 
ABSTRACTING AND INDEXING SOCIETIES Gon. !renee . . 17 
ISTITUTO PER GLI STUOI SEMANTICX E COGNITIVI . . . . 18 
PETRARCH Translated by computer , . . . . . . . . . 20 
STENOTYPE Computer-aided transcription . . . . . 21 
5ERIALS LIBRARIANSHIP Drexel Library c)uarterly , . . 22 
A COMPUTER SIMULATfON OF AMERICAN SIGN LANGUAGE . . 23 
PLATON--A NEW PROGRAMMING LANGUAGE FOR NATURAL 
LANGUAGE ANALYSIS Makoto Nagao & Jun-Ichi Tsuj Li 28 
I. Introduction . . . . . . . . . . . . . . . . 28 
11. Pattern-matching . . . . . . . . . . . . . . 31 
III. Basic operations of PLATON . . . . . . ... 34 
IV . Push-down and pop-up operations . . . . . . . 40 
V . A sl-mple example . . . . . . . . , ... 44 
VI , Conc lus ion . . . . , . . . . . . . . . '8. 52 
ACL MEMBERSHIP DATA - . . . * - . . . . . - 54 
EDITOR'SREPORT . . . . . . . . . . . . . . . . . . 55 
SURVEY OF MEMBERS Preliminary report . . . . . . . 57 
IEEE TUTORIALS. . . . . . . . . . . . . . . . . . . 81 
American Journal of Computational Linguistics 
Microfiche 37 : 3 
BIOGRAPHICAL NOTE 
STANLEY R, PETRICK 
Dr. Petrick works on syntactic and semantic analysis for 
transformational grammars; he has been employed in this field 
since 1967 as a member of the Theoretical and Comoutational 
Linguistics Group at the IBM Thomas J. Watson Research Center 
His doctoral dissertation (Linguistics, MIT, 1965) was A 
Recognitim Procedure for Transformational Grammars; it was 
one of the original systems to go beyond phrase structure. 
Dr. Petrick programmed the Ordvac Computer in 1953 and went 
on to study digital computation at MIT as a student officer 
in the Air Force. From 1958 to 1967, the was a research 
mathernetician in the Applied Mathematics Branch ~f the Air 
Force Cambridge Research Laboratory. His research was on 
truth function minimization, speech recognition, and formal 
language theory, 
He was chairman of the ACM Special Interest Group on Symbolic 
and Algebraic Manipulation, 1969-1971, program chairman of the 
1968 ACL meeting, and vice president of ACL in 1975. 
American Journal of Computational Linguistics 
Microfiche 37 : 4 
PERSONAL MOTES 
BRUDERER HERBERT TO Finkenweg 3, 3110 Munsingen/Berne, 
from Haslerstrasse 12, Berne. 
DE MEY. MARC T. M. TO Department of Psychology, University of 
Tilburg, Hogeschoollaan 225, Tilburg, Netherlands, from 
Gent, Belgium. 
DOBREE. NICK To Rose Cottage, Hindon, Wiltshire, England, from 
Beirut, pending the cessation of hostilities. 
RAMEH, CLEA To School of Languages and Linguistics, George- 
town University, from a year in Brazil. Rameh is chairman 
of the 1976 Round Table on Languages and Linguistics. 
REDDY. D. RAJ Guggenheim fellowship for studies in the com- 
puter processing of speech. 
WALTER. JOHN PAUL TO 1746 Curtis Avenue, Manhattan Beach, 
California 90266 
American Journal of Computational Linguistics 
~i crofiche 37 : 5 
GEORGETOWN UNIVERSITY ROUND TABLE 
LANGUAGES AND LINGUISTICS 19 76 
' 'Semantics: Theory and Application' ' 
March 11, 12, 13, 1976 
The theme of the Georgetown University on Thursday evening, on Friday mcr.rine and after- 
Round Table on Languages and Linguistics 1976 noon, and on Saturday morning, March 1 I-? 3 On 
is semantics. There will br two types of meetings: Friday esening. Georgetown University will hast a 
(1 ) fourtcen Interest Group Sessions on Thursday budet reception for those attending the Round Table. 
afternoon, March 1 1; and ( 2 ) four Plenary Sessions 
The Interest Group Sessions will provide an oppor- 
tunicy for informal discussion of various related topics. 
These sessions will be chaired and their topics wjli be in- 
troduced by scholars in the various fields. A limit of 25 
participants for each Group should provide a manageable 
interchange of iieas. Acceptance for each group will depend 
on early preregistratim. On the form below participants 
should indicate their preferred Interest Group and list any 
speciiic topic they desire did. Those not registering 
until the Round Table itself cannot be guaranteed par- 
ricipacion in an Interest Group. 
The Plenary Sessions will explore the relationship of 
semantics to various theoretical viewpoints in linguistics, 
to problems in applied linguistics and to problems in 
language teaching. The Plenary Sessions will also discuss the 
relationships of semantics to anthropology, computer sci- 
ence, philosophy, and psychology. 
Interest Group Sessions / March 11, 1976 
TOPlCS AND CHAIRMEN 
1 TEFL: Englisb for Special P~rposes 
James Alatis, Georgetown University 
2 Commuaication Slrategies md Modem Lagutage 
Teaching 
Frc 'nric Boxo, Georgetown University 
3 Tbe Semariics of Tense and Aspect in English 
Walter A. Cook, S.J , Gecrrgecown University 
4 Semantics, Cognition, and ibe Brnin 
William Orr Dingwall, University of Wryland 
5 Linguistics, P~~chorrrtdysis, and Dylramics of Lan- 
guage in Groups 
Robert DiPietro, Georgerow n University 
6 Smantics ad dc Teaching of English to Speakers 
of Otbcr Didects 
Ralph Fasold, Georgetown University 
7 Litrguistics ad Early Readitzg 
Robert Lado, Georgetown University 
8 Senzantics ad Conzprrter Science 
R. Ross Mncdonald, Georgetown University 
9 Literattrre for Basic Lalrguage Skills 
Marzieh Samii, Ferdowsi ( Meshed) University 
I0 Acquiring MeutaitPg in s Secod Cuhure 
Muriel Saville-Troike, Georgetown Univer~ity 
I 1 Historical Semantics 
Shaligram Shukla, Georgetown Univefsicy 
1 2 Sign Latrguage, Semantics, Semiotics 
William C. Scokoe, Gallaudet College 
1 3 Bilingual Educatf on 
F. LeRoy Walser, Office of Bilingual Education, USOE 
14 Comprehension in &ding 
Rose Marie Weber, Cornell University 
Plenary Sessions / March I I - r 3, r 976 
HALL OF~NATIONS, WALSII BUILDING, 36~~ AND N STREETS, N.W. 
March ll, 7: 30-1O:OO p.m. March 12,2:00-5:00 p.m. 
Sentattics: Some Synchtotaic d Diachronic Aspact$ Set~nticr md Otbn Discipli#es 
Dwight Bolinger, Palo Alto, Califom ia Oswald Werner, Northwestern University 
Winfred P. Lehmann, University of Texas, Austin Charles Rieger, Univcrslty of Maryland 
Gilbert Hnrnian, Princeton University 
Nfomo Caramad, Johns Hopkjns University and 
Concordin University 
March 12,9:00--12:OO noon hch 13,7:00-12:OO noon 
Setmntios and Ling~distics S~ntmtics *nd hguagfi Pdagogy 
Jorge E. Hankamer, Harvard University Frances M. Aid, Florida International University 
Kay S. Jackendoff, Brandeis University Milton M, Azevedq University of &!orado 
Geoffrey Leech, Univers~ty of Lancaster Christopher N. Candlin, University of Iancaster 
Barbara Hall Partee, University of Massachusetts Ragnhild Soderberg, University of Stockholm 
GEORGETOWN UNIVERSITY 
SCHOOL OF LANGUAGES AND LINGUISTICS 
Preregistration for the Round Table will 
NAME: 
dose March 8. The preregistration fee is LAST (pleare prinu or iype) FIRST MIDDLE INITIAL 
$6.00; the registration fee is $8.00. Early 
ADDRESS: 
preregistration should guarantee participation 
in the Interest Group desired, but since par- 
(z+ code) 
ticipation is limited, preregis&ants should in- 
dicate the order of their preference and list 
PROFESSIONAL AFFILIATION: 
INTEREST GRCYl.JP 
any specific topic they desire discussed. 
PREFERENCE: (1) -- I (2) -, (3), -- (4; -- 
SUGGESTIONS FOR TOPI.CS OF DISCUSSION: 
Make chechr puyable to GEORGETOWN UNNERSITY. 
MAIL TO: Clea Rameh, Chairnta}~ 
Georgetown University Round Table 
on Languages and Linguiscra 1976 
GEORGETOWN UNIVERSITY 
WASHINGTON, D.C. 20057 
Telephone: (202) 625-302 1 
American Journal of Computational Linguistics ~i crofiche 37 : 7 
Science and Our Expectations: 
The Reach and the Grasp 
Annual Meeting 
Boston 
18-24 February 1976 
PROGRAM EXCERPTS 
EXTRATERRESTRIAL INTELLIGENCE 
Extrasolar systems, origins of life, Ozma search, beacons, 
Soviet searches, extragalactic systems, long-range strategies. 
THE INTEGRATION OF THE NATURAL AND SOCIAL SCIENCES 
Philosophy, humanism, normative factors 
EDGAR TASCHDJIAN, ERVIN LAZLO, ALFRED KUHN 
TELECOMMUNICATION, TRANSPORTATION. AND URBAN DEVELOPMENT 
Substitutions for travel, urban growth, alternatives. 
AAAS MEETING 
Automatic speech, computer programing, offices, home terminals, 
brain-computer hookups. 
THE PERISHING PUBLISHING PROSPECT FOR SCIENTIFIC AUTHORS 
Economics and a resourceful technology are eroding the comfor- 
table system by which post-World War I1 scientists have gained 
credit for results of their research. Existing arrangements 
for citation and indexing for reference and retrieval are simi- 
larly threatened. This Symposium will attempt to alert scien- 
tific authors to trends, opportunities, and dangers they must 
face in the next decade, as foreseen by knowledgable communica- 
tors with experties and occupational mandates to define and 
predict the future. Emphasis in all presentations will be on 
the impact of imminent changes on individual authors rather than 
on specialized problems in technical and scientific information 
transf ern 
ARRANGED BY HAROLD F. ~SBORNE. ASSISTANT FOR INFORMATION, 
DEPARTMENT OF MEDICINE AND SURGERY, VETERANS ADMINISTRATION 
SCIENTIFIC JOURNALSa AN ENDANGERED SPECIES 
Robert A. Day, Managing Editor, American Society for Microbiology 
THE PUBLISHING OUTLOOK FOR 1985 AND BEYOND 
Robert A. Harte, Executive Officer , American Society of Biological 
Chemists 
THE CITATION OUTLOOK FOR 1985 AND BEYOND 
Ben H. Weil, President, National Federation of Abstracting and 
Indexing Services 
TRENDS IN TFC!+NOtOGICAL INNOVATION FOR SCIENTIFIC COMMUNICATION 
Seldci~ W. Terrant, Head, R&D, Books and Journals Division, 
American Chemical Society 
ELECTRONIC ALTERNATIVES TO PAPER-BASED COMMUNICATION 
H. E. Barnford, Program Director for Access Improvement, Off ice 
of Science Information Service, National Science Foundation 
AAAS MEETING 
Information sector, information as a commodity, location deci- 
sions, productivity, international trade, market aspects, mass 
production 
SCIENTIFIC COMMUNICATIONS AND THE ADVANCEMENT OF SCIENCE 
Information is both the input of raw rraterial and the output of 
scientific and engineering activities. Comunica.tions services 
are the means by which scientists and engineers are able to 
build upon the results of prior work even though that work may 
have been done at a distant place or in a different discipline. 
Many prestigious committees and study groups have analyzed the 
information problems of scientists and engineers and offered a 
broad range of recommendations for improvement. But very little 
has changed. This symposium will review the problems with, and 
needs for, co~unications services by the scientific community 
and suggest new approaches to the problem of maintaining an eco- 
nomically viable and user responsive scientific and technical 
comunications enterprise in the U.S. Speakers will deal with 
a variety of specific questions including: What is the role of 
information in science and engineering? How much progress have 
we made in improving information services during the last 10 
years? What are the current problems in financial support, or- 
ganization, and management of the scientific communication enter- 
prise? What can be done to improve the extisting situation and 
enhance the value of scientific and technical information ser- 
vices to users engaged in scientific research? 
ARRANGED BY RUTH M. DAVIS, DIRECTOR, INSTITUTE FOR COMPUTER 
SCIENCES AND TECHNOLOGY, NATIONAL BUREAU OF STANDARDS, AND 
JOEL 0. GOLDHAR, PROGRAM DIRECTOR FOR USER REQUIREMENTS, OFFICE 
OF SCIENCE INFORMATION SERVICE, NATIONAL SCIENCE FOUNDATION 
PROGRESS IN IMPROVING COMMUNICATIONS AMONG SCIENTISTS 
Alvin M. Weinberg, Director, Institute for Energy Analysis 
THE ECONOMIC CONDITION OF THE SCIENTIFIC COMMUNICATIONS ENTER- 
PR I SE 
Fritz Machlup, Professor of Economics, New York University 
THE ROLE AND IMPORTANCE OF COMMUNICATIONS IN THE ADVANCEMENT OF 
SCIENCE 
J. Herbert Hollomon, Director,, Center for Policy Alternatives, MIT 
AAAS MEETING 
COMMUNICATING ABOUT SCIENCE TO THE PUBLIC 
Edwin D. Canham, Editor Emeritus, Christian Science Monitor 
NEXT STEPS--A PLAN OF ACTION FOR IMPROVING COMMUNICATIONS 
SERVICES IN SCIENCE AND ENGINEERING 
Lewis M. Branscomb, Chief Scientist, IBM Corporation 
COMPUTERS AND MEDICAL KNOWLEDGE: EXTENDING THE AVAILABILITY 
AND USE OF KNOWLEDGE IN MEDICINE 
The symposium explores new directions of development extending 
the use of medical knowledge via c~mputer-based systems. In 
one category of such systems, knowledge is extracted automati- 
cally from the aggregate of patient records (medical data banks); 
another category encapsulates and makes available decision 
making rules of human experts. The program describes the cur- 
rent state of the art in the development of these systems; 
reviews research in the modeling of clinical decision processes; 
examines the implications of computer-aided consulting on medi- 
cal practice and education; and surveys Federal programs which 
support the training of manpower for biomedical information pro- 
cessing 
ARRANGED BY VLADIMIR SLAMECKA, DIRECTOR, SCHOOL OF INFORMATION 
AND COMPUTER SCIENCE. GEORGIA INSTITUTE OF TECHNOLOGY 
USE OF COMPUTERIZED DATA BASES TO AID IN PATIENT MANAGE h1 
DECISIONS 
R. A. Rosati, Assistant Professor, Duke University Medical Centel 
A PERSPECTIVE ON KNOWLEDGE BASED DECISION SUPPORT SYSTEMS FOR 
CLINICAL MEDICINE 
Harry Pople, Co-Director, Medical Information System Lab., Pitt. 
METHODOLOGIES OF RESEARCH IN CLINICAL PROBLEM SOLVI 
Albert N. Radre, Information arid Computer Science, gia Tech. 
IMPLICATIONS OF COMPUTER-AIDED MEDICAL CONSULTING 
Ralph L. Engle, Jr., Medicine and Public Health, Cornell University 
THE EDUCATION OF BIOMEDICAL INFORMATION/COMPUTER SCIENTISTS 
AND PROFESSIONALS: FEDERAL TRAINING PROGRAM 
Roger W. Dahlen, Chief, Division of Biomedical Medicine, NIH 
American Journal of Computational Linguistics Microfiche 37 : 11 
APPLICATION DEADLINES 
NATIONAL ENDOWMENT FORTHEHUVIANITIES 
MARCH -- OCTOBER, 1976 
FELLOWSHIPS 
Summer Seminars for College Teachers, 1976 
RESEARCH GRANTS 
Centers of Research, beginning after December 1, 1976 
EDUCATION PROGRAMS 
Education Proj ect Grants, beginning after December 1 
EDUCATION PROGRAMS 
Consultants Grants (beg. 6/15); Planning Grants (L2/1) 
YOUTHGRANTS 
Projects beginning after October 1, 1976 
RESEARCH GRANTS 
Research Tools and Editing, beginning after Jan. 1, 1977 
PUBLI c PROGRAMS 
Projects beginning after October 1, 1976 
FELLOWSHIPS 
Fellowships for Independent Study and Research, 1977-78 
RESEARCH GRANTS 
General Research, beginning after January 1, 1977 
EDUCATION PROGRAMS 
Program Grants (177) Consultants Grants (9/15/76) 
EDUCATION PROGRAMS 
Development Grants, beginning after' May 1, 1977 
PUBLIC PROGRAMS 
Projects beginning after January 1, 1977 
EDUCATION PROGRAMS 
Consultants Grants, beginning after December 15, 1976 
RESEARCH GRANTS 
Centers of Research, beginning after July 1, 1977 
American Journal of Computational Linguistics 
Microfiche 37 : 12 
1976 National Computer t?onference, 
76 NCC. c o AFIPS. 210 Summit Ave 
The scientist has prerogatives, duties, and obligations in 
government, and so do professional societies. Anthony Ralston, 
president of AFIPS and past president of ACM, will speak on 
this theme at a plenary session of the 1976 NCC. 
Federal coxmnissions an privacy and electronic fund-transfer 
systems are among the developments he cites that raise ques- 
tions of professionalism, codes of ethics, good conduct, 
measures of proficiency, public protection, and social and 
political awareness. 
Dr. Ralston is professor and chairman of Computer Science in 
the State University New York Buff a10 . 
American Journal of Computational Linguistics 
Mi crofi che 37 : i 3 
SECOND WORKSHOP ON COMPUTER ARCHITECTURE 
FOR NON-NUMERIC PROCESSING 
GAINESVILLE, FLORIDA, JANUARY 22-23, 1976 
SPONSORED BY ACM SIGIR, SIGMOD, SIGARCH 
Stanley Y. W. Su and G. Jack Lipovski, Co-Chairmen 
PROGRAM 
FILE OPERATIONS IN A STREAMING PROCESSOR 
B. W. Jordan, Jr., Northwestern University 
K. J. King, Digital Equipment Corporation 
G. D. Miller, Bell Telephone Laboratories 
ARCHITECTURE OF A NONNUMERICAL ACCUMULATOR 
Yaohan Chu, University of Maryland 
HIERARCHICAL MICROPROCESSOR ORGANIZATION 
David R. Smith, SUNY, Stony Brook 
A GENERAL APPROACH TO FUNCTIONALLY DISTRIBUTED COMPUTER ARCHITECTURE 
A. Reszka, Teletype Corporation 
M. J. Gonzalez, Jr., Northwestern University 
STUDY OF MACHINE ARCHITECTURES FOR SPECIALIZED INFORMATION 
RETRI €VAL COMPUTERS 
B. J. Hurley and Duncan Lawrie, University of Illinois, Urbana 
A SPECIALIZED COMPUTER ARCHITECTURE FOR HIGH-SPEED TEXT SEARCHING 
David C. Roberts, Central Intelligence Agency 
AN ARCH1TECTUR.E FOR THE EFFICIENT COMBINING OF LINEARLY ORDERED 
LISTS 
Lee A. Hollaar, University of Illinois, Urbana 
A MULTIPLE ASSOCIATIVE MEMORY ORGANIZATION FOR PIPELINING A 
DIRECTORY TO A VERY LARGE DATA BASE 
P. Bruce Berra and Ashok K. Singhania, Syracuse University 
WORKSHOP ON COMPUTER ARCHITECTURE 
A TWO-LEVEL ARCHITECTURE FOR A LARGE DATA BASE 
Tomas Lang, University of California, Los Angeles 
Ezatollah Nahouraii, IBM 
PERFORMANCE EVALUATION OF A RELATIONAL ASSOCIATIVE PROCESSOR 
E. A. Ozkarahan, S. A. Shuster, and K. C. Sevcik 
University of Toronto 
A VIRTUAL MEMORY SYSTEM FOR A RELATIONAL ASSOCIATIVE PROCESSOR 
S. A. Schuster, E. A. Ozkarahan, and K. C. Smith 
University of Toronto 
IMPLEMENTAT I ON OF A CONTEXT-ADDRESSED  PIPELINE^ S1bd-D ARCHITECTURE 
S. J. Ackerman, A. Eman, G. J. Lipovski, and S. W. Su 
University of Florida 
SOFTWARE ASPECT OF THE CASSM SYSTEM 
Atrmed Ernam, StanLey Y. W. Su, and G. J. Lipovski 
University of Florida 
American Journal of Computational Linguistics 
~i crofiche 37 : 15 
AYERICAM SOCIETY FOR INFORMATION SCIENCE 
The deputy direccor of the National Library of Medicine was 
formerly head of the Office of Science Information Service, 
National Science Foundation. 
He was chairman of the White House Committee on Scientific 
and Technical Information; is chairman of the Federal Library 
Committee s executive committee, a member of the UNISIST 
advisory committee, and chairman of the Panel on Environmental 
Quality Information of the Organization for Economic Cooperation 
and Develo~ment 
American Journal of Computational Linguistics ~jcrofichc 37 : 16 
ANNUAL MEETING OF THE ASSOCIATION FOR 
LITERARY & LINGUISTIC COMPUTING 
UNIVERSITY OF AMSTERDAM 
DECEMEIER 13, 1975 
Computer-controlled sampling for bilanguage dictionary compilation 
R. D. BATHURST, UNITED KINGDOM 
Les activites du Lexique Intellectuel Europeen - traitement 
electronique des textes d'auteurs 
T. GREGORY, ITALY 
On lexicographical computing - some aspects of the work for a 
Mexican SpanLsh dictionary 
I. F. LARA, MEXICO 
The recognition of finite verbs in French texts 
BENTE MAEGAARD, DENMARK 
Address 
American Journal of Computational Linguistics Microfiche 37 : 17 
NATIONAL FEDERATION OF ABSTRACTING AND INDEXING SOCI ETJES 
INFORMATION 
Dilemmas, Decisions, Directions 
WEDNESDAY MARCH 10, 1976 
30 A.M. 
2:00 P.M. - REGISTRATION 
TUESDAY MARCH 9, 1976 THEME SESSION II I: Indexing in Interactive 
HEME SESSION I: Current Research Projects 
Related to Abstracting and lndexing 
9:00 A.M. - 1.1:45 A.M. 
Cha~rrnan Cec G. Burchival (NSF OSIS) 
Sess~or) Note: The ci~vclo(~mt:r~t of new tcchn:)f 
ogles has createcl th~! pcterit~i11 ~OI ~mprovemc:r>ts 
~ri ~r~forrnat~on svstems. Tht! NSF OSlS IMS taken 
a Icadlrir~ role 111 supl)ort~n!; I tbsc?arch to contr~t)utr~ 
to ~ncrcasctl r~atloclill ~~rocluct~wrly through the: IISC 
of sercnr~f~c and techrlrcal rr~forrnar~orl Ht?ports of 
some current O~ISIC: r(!sc;:t.:: I)roj(!cts r(!/3[(!0 10 
al)stracl~rrq ~~ritl ~rl(it!x~ri(j WIII I)u ~)rt:scrltt!(l. 
LUNCH 
lAfle~)fh!~!s to rr~ilko ow11 ;~rri)~~r)c:~nt~i~Is) 
HEME SESSION I I: Changing Patterns of Primary 
Sources 2:30 P.M. 5:00 P.M. 
c/tillf f?~ifll A. 146)0(1 fJOt)(:rf~ [(:A[-) 
St!sslo~l Noro Not orlly IS t1i1: tl11;11\111 y r)f In.Irllilry 
tlt(!rilt~llf~ ~t!(~l~lt'l~(~ lll(jt!~lll(] ilfl(1 i~t)slrill:flrNJ lrl 
crc?asirlg, I~trt ZIISO thr? format of tht? pr Irnilry sources 
is diariqrrr~j. Tllt:sc! rrcw fc)r.milts ~nclu(lc flit 
review journal, tht: synootlc journal, microforms 
and videotape. These new clnections have a ciefiri 
Ite Impact on the abstractllig ;and ~ndux~nq corn 
munity. 
5:OO P.M. - 6:00 P.M. 
FEDERATION ASSEMBLY 
BUSINESS MEET1 NG 
6:00 P.M. - 9:00 P.M. 
CONFERENCE-WI'DE RECEPTION 
at 
Tho Center of Science and Industry 
Sponsored by Cficmical Abstracts Service 
Systems -- 9:00 A.M. 
11 :45 A.M 
Chairmarl Marv~n Wilson (NTIS) 
Session Note Inforrnat~orl disserninatlon centers 
irttlrze hath data t~ases created by the abstiiacting 
and indexing community and software created by 
the center Ttiir combination of resources creates 
t~oth problems and. enrlchme~t in the transfer of 
~nformat~on in the intcractivc mode. The identi- 
f~catlor~ ot thest! clilernmas will promote new dls- 
cusslor1 ancl (11 rcxt lolls for ~rnl)rovcrl 141ssc.minat1on 
of ~r~forrnat~ori. 
MILES CONRAD MEMORIAL LECTURE 
Presented by Frederick G. Kilgour (OCLC) 
CONFERENCE LUNCHEON 
12:30 P.M. - 2:00 P.M. 
THEME SESSION IV; New Dircctions in User 
Education - 2:00 P.M. - 4:30 P.M. 
Chairmarl Roger K. Summit (Lockhced Aircraft 
Cor 11.) 
Session Note: User eciucat~on involves many sty- 
nlcnts of the ir~forrnatron community and often 
lt:ads to fraqrneri tation of the knowlcclgc! r~ccctecl to 
utilize! the resources ar1r.i services avai lal~lc. Real 
nccds rnttsl brr iclentific?d and cc,n!r?rlt ~>rc,viclaci in 
rrcw morics t hill c:rlal)l(: t tie rnultil)l,s irst:rs ~n t hc 
irlformatiorl ~:tiii~~l to 11ii~xir1ii7tt Ilruir unc.lcrslaricl~r~g 
01 malrrrrals provrtlrxt. 
American J ournaI of Cornputat ional Linguistics Microfiche 37 .- 18 
-- BIBLIOGRAPHY 
ISTITUTO PER GLI STUDI SEMANTICI E COGNITIVI 
FONDAZIONE DALLE MOLLE 
17 RUE DE CANDOLLE 
1205 GENEVE I SWITZERLAND 
The new address is in effect since January 1, 1976. 
A list of reports was printed on Microfiche 22:46. Abstracts 
can be located through the index. 
18. Semantics, preference, and inference. A full description 
of a system and a program. To appear. 
19. Frames, planes, and nets. Greg Scragg. SF40 HC, SF5 MF. 
20. A structure for actions. Greg Scragg. SF40 HC, SF5 MF'. 
21. On understanding German noun clusters. dolfgang Samlowski 
SF 20 HC, SF2.50 MF. 
22. A brief on case. Eugene Charniak. SF30 HC, SF 4 MF. 
23. A process to implement some word-sense disambiguations. 
Philip Hayes. February 1976. 
24. On the referential attributive distinction. Eugene Charniak. 
February 1976. 
ISTITUTO PER GLI STUDI SEMANTICI E COGNITIVI 
25. Several ways to be suggestive. Margaret King. February 1976. 
26. Pragmatic aspects of noun cluster understanding in German. 
SF20 NC, SF2.50 MF. 
Prices of older reports in Swiss Francs: 
No HC MF 
1 5.00 
3 2.50 
4 LO. 00 
5 4.00 
6. 2.50 
7. 5.00 2.50 
8. 5.00 2.00 
9. 5.00 2.00 
1.0. 5.00 4.00 
11. 4.00 
12. 5.00 
13. 2.50 
14. 2.50 
25.00 Proceedings of the Tutorial on Computational 
Semantics 
The Rcrum Familiarium tibrt corn. 
prises about 350 Latin prose letters 
American Journal of Computational Linguir tics 
written by Franccsco Petrarch to a 
numhcr of corrcspondents: classical, 
contenlporary, and some even fic- 
tit ious. Dr. Bernardo's translation was 
recently published by State University 
of NFW York Press. 
The tr;anslation began in carly 1969 
when tl~e definl'tivc edition of thc Latin 
Pet rarc h Translated by Computer ,, ,o,s~ ,,, ,,sc, ,,, 
- - 
keyppnched with pagination and para- 
SUNY Binghamton linguist finds computer a malor 
graphing into Binghamton's corn- 
time-saver in work on Petrarch's early letters 
puter. By that summer an alphabetical 
word list of unique forms was ready to 
he translated into English. 
A work some consider to be es- 
Italian and comparative literature at 
With a team of five Latinists, Dr. 
sential to a firm grasp of Humanism 
SUNY Binghamton. The project was Bernardo entered as many as seven 
and the Renaissance, the cultural 
made possible through funding from meanings for each form on speciaI 
movements that mark the beginning of the SUNY /Research Foundation printout sheets. The alphabetical list- 
modern times, has been translated 
Awards Program, and with the co- ing was then converted into a chrono- 
with the assistance of the computer by 
operation of Binghamtoh's Computer logical ordering of the forms, starting 
Dr. Aldo S. Re.mrdo professor of Center. 
with the first word of the first letter. 
Using these he then translated the first 
eigh;books, or 144 letters, which this 
I volume includes. 
A concordance of the entire col 
lection was prod~ced as a sort of by- 
product. This is to be published by 
SUNY Press shortly, according to Dr. 
Bernardo. 
Dr. Bemardo believcs the campuler 
is a great time-saver in translat~ons of 
this kind. Never have we had in- 
dimvidual Renaissance Latin forms in so 
much context, with fivc ur six Latin 
words before arrd after each kcy word. 
The very lnetcr of Kcnilissante Latin 
car1 be studied now bcbcausc of the 
llugc corltc8xt thcse pririrouts oFft-r," 
hr siiys. 
A n~rdol. tr~idrrtilking in humanis~ic 
rcsc:ircli tllc trnnslalion is tllc first 
altrl1l pt in Errglish to nlakc 3vailiil~Ic* 
; all of I'ctrilrc h s eai.lies[ slid pcrllaps 
nicat in~parra nt col'lcctiorl of ~)rc~se 
COMPUTER PRINTOUTS were a constant reference suuwc as 
Dr. Aldo S. Bcrnardo of SUNY Bingllarnron dictated his rra~rslalion 
of Pcrrafcb s "Rcrum Fawiliarium Libri, a cdlcction oi some 350 
letters written by the famous Rcnaissance port to a varic~y uf corrcspondrnts. 
The work has ken published by thc SUNY Press. 
From Chronica, Published by the Research 
lcttcrs. 
l'hc*. St JNY Kcsc~arcl~ Founda~ion 
Awards granted llr . Bt-rnarcio in- 
cluded a 815,300'Joitlt Awards Council 
Award in 1967 (under a special pro- 
gram lastirlg for only one vcar) and a 
$3200 combination grant.in aid and 
fellowship awarded in 1969. 
~oundation of the State Univer'stiy of 
New York 
Hebc~ ttttlciry 
Public Infor~l~atmn Office 
SUNY Bingllam\ot~ 
American Journal of Computational Linguist ier 
Mi croficl~e 37 : 21 
Stentran Systems, Inc , located at 380 Maple Avenue, West, in 
Vienna, Virginia 22180, has "developed a program for the com- 
puter translation of stenotype into English,'' according to 
Michael E. LaBarde, Director of Linguistics. Experience with 
their system was described in two 1975 issues of the publica- 
tion of the National Shorthand Reporters Association by Gilbert 
Frank Halasz. The Association has a Committee on Computer- 
Aided Transcription; their chairman is Doris 0. Wong, 30 Milk 
Street, Boston 02109. 
A news release from Stectran, dated 1975, states that their 
system can nrovide, besides the transcription, "an overview of 
the information arranged in conceptual fashion" 
"After analysis of the semantic or legal road map which has 
been prepared for him," the release goes on, "the attorney is 
then in a position to demand additional litigation support 
from Stentran in the form of key words, phrases, dates, con- 
junctive recall of concepts or ideas, or any verbal patterns 
which he feels can asist him in reaching legal conclusions. I! 
American Journal of Computational Linguistics 
Microfiche 37 : 22 
DREXEL I.l"URRY QUARTERLY COVERS CURREII'T ISSUES 1 N SERI4LS L I DP\ARIA?ij!l! P 
..I... ---.----.--.-.--- - ---. . .. * -II.-__ _ _ 
------ -.- 
PHiLACELPHIA--The Drexel Library Quarterly, Volume 11 , no. 3, e~aeinei 
- 
"Current issues in Seris i; l ibrari~rlship. " Serials 1 ibrarians ofte 
difficulty ident 
to their work. 
are affecting on- the-job ! I 
The a~ticies , selected for their tinel iness , often cmphasi ze the r;;anner 
in \vhich sarials librarians can have impact upon the issues un 
Oeni ta 
Meber, Serials Librarian of Nontgomery County (pa. ) Con~uni ty 
College, and Toni Carbn Bearvan, Executive Director of the Natiocal 
Federation of Abstracttirig and Indexing Services, are guest editors for the issua 
Articles included in the iswe are: "The Serials Librarian As Activist" 
It I- I. 
by David C. Taylor; Itain cn~ry for Serials" by Joseph J. H~!qard, "ISR[I.!S) 
and Title Main Entry for Serials" by C. Sumner Spalding; "International 
b 
Cooperation in Serials" by Jos~oh !4. I'rice; "Hatiorla1 Sclris 
I I 
by Mary Sauer; "The CO#SER Project." by Paul Y assall o; "COASER Int-br-Re1 a-ti onships" 
by Lawrence G. 'Seriql:; : Costs and Budget Projections" hy 
F.F. Clasquin; and "Education of Serials Librariansu 
Copies of Vol . 11, no Serials Librarianship," are 
available for $4 each ( obtside the iJSd ::nd Canad?) froin the Orexel Library 
Quarterly, Graduate Schou l of li brary Sci crlce, 
Philadelphia, Pa. 19104. Telephone: 21 5-895-2 
American Journal if Computational- Linguistics 
Microfiche 37 : 23 
HARRY W. HOEMANN, VICKI A. FLORIAN, AND SHIRLEY A. HOEMANN 
Bowling Green State University 
Ohio 43403 
Stokoe (1960) has designed a model of the structure of the 
American Sign Language (ASL) which is amenable to computer si- 
mulation. He has proposed that signs comprising the ASL lexi- 
con are composed of three basic aspects, location (TAB), hand 
configuration I (DEZ) , and movement (SIG) . He has identified a 
finite number of each of these elements, and he has proposed 
that they may be combined in various ways to constitute recog- 
nizable and meaningful signs. Recent reformulations have ven- 
tured some modifications (e.g., Stokoe, 1972), but the basic 
approach remains the same. 
Such a conceptualization of ASL implies that if a computer were 
furnished a set of each of these types of elements, it ought to 
Acknowledgments. This investigation was supported by NIH Re- 
search Grant NS-09590-05 from the National Institute of Neuro- 
logical Diseases and Stroke. We thank the J. Preston Levis 
Regional Computer Center, Perrysburg, Ohio, and Charles M. 
Bernstein, Bowling Green State University, for assistance. 
AMERICAN SIGN LANGUAGE Microfiche 37 : 24 
be able to compile the signs that are composed of the features 
in its repertoire It needs to be emphasized that such a com- 
puter simulation is not merely a matter of cartooning, although 
this lies within the capability of computer graphics to portray. 
METHOD 
APPARATUS. Our principal apparatus was an Owens-Illinois 
Digivue Plasma Display Unit attached to a Nova 1220 minicomputer 
The Digivue Display Unit is an electronic device with a gas- 
filled display matrix and activating circuitry. An electrical 
signal is passed through each of two very fine wires at right 
angles to one another. When the gas is activated, it lights up 
at the poiat of intersection. There are 512 grid lines in each 
direction, making a total of 262,144 addressable points. Points 
can be written or erased at the rate of 50,000 dots per second. 
The program language used was Graphic Basic, a version of Data 
General Corporation's time-shared BASIC, modified for use with 
the Digivue Display Unit (Fulton, 1974). Access to the computer 
was gained through the Digivue keyboard and through a teletype- 
writer. Paper tape output from the teletypewriter provided 
permanent storage of previously written programs. 
PROCEDURE. AS a concession to storage limitations, the project 
was, at first, limited to one-handed signs located on or near 
the face. An oval shape with stylized browline, nose, and 
mouth was stored in memory and served as the reference for any 
sign which the computer was required to generate 
Hand co~~figurations (DEZ) were constructed by joining coordinate 
points with line segments. The points were stored in the memory 
of the computer, and graphic commands were issued to form the 
line segments. 
As a program was fed into the computer, it stored in memory in- 
formation about the DEZ and the initial TAB while the face was 
drawn on the screen. Another series of subroutines drew the 
DEZ in its initial TAB and moved it to another TAB so as to re- 
present a movement (SIG). If the sign required a change of DEZ, 
the program could call the new DEZ up from memory and place it 
in the sign's final position. 
VALIDATION. The adequacy of the simulated signs rests with 
their intelligibility. Deaf and hearing persons who were fluent 
users of ASL were tested to verify that the graphic display 
yielded signs that could be recognized as part of the ASL lexicon. 
RESULTS AND DISCUSSION 
Thus far 11 DEZ, 8 TABs, and 7 SIGs have been programed and 
stored on a single paper tape. If disc storage and the necessary 
interface were added to our Nova, it is likely that we could 
represent all 12 TABs, 19 DEZ, and 24 SIGs identified by Stokoe 
as primes of his model. Meanwhile, the subset of structural 
features programed thus far constitute a suitable feasibility 
study of a computer simulation of ASL. 
The following results 
have been achieved 
It has been verified that the features of ASL identified by 
Stokoe as structural elements may also function as distinctive 
features. The same DES and SIG executed with different TABs 
may result in signs with different meanings (father and mother, 
summer and dry in our data). The same TAB and SIG executed with 
different DEZ may also result in signs with different meanings 
(who and lipread in our data). Finally, the same DEZ and TAB 
with different SIGs may also result in signs with different 
meanings (summer and wise, face and who in our data) . 
Secondly, if any one of these aspects (TAB, DEZ, or SIG) of a 
sign is altered, the resulting sign compiled by the computer is 
1) 
likely to be nonsense" in ASL. In our data, the 11 DEZ, 8 
TABs, and 7 SIGs yield potentially 11 x 8 x 7 or 616 signs. 
Over 600 of them are nonsense. This indicates that most signs 
in ASL differ from one another in more than one distinctive fea- 
ture. Also, since many signs change DEZ and involve more than 
one SIG, there is a low probability of confusing one sign with 
another, even when the signs are presented out of context. 
Finally, our data indicate that the orientation of the hands 
also constitutes a distinctive feature in ASL. Inappropriate 
hand orientation can disrupt intelligibility even when the other 
three aspects are compiled correctly. 
Stokoe's model seems to 
be sufficiently robust to assimilate the required revision with- 
out altering his basic approach. 
This computer simulation of ASL was limited to the structural 
features of individual signs, and it corresponds to a study of 
the phonological structure of spoken languages. (Stokoe refers 
to his analysis as CHEROLOGY after the Greek chcir or hand .) No 
attempt was made in this simulation to present the signs in a 
linguistic context or to represent the structure of AS sentences 
Future studies are planned in which the graphic di~play of indi- 
vidual signs will be subject to systematic distortion to discover 
whether TAB or SIG aspects of signs are perceived categorically 
by native users of ASL. 
American Journal of Computational Linguistics Microfiche 37 28 
MAKOTO NAGAO AND JUN-ICHI TSUJII 
Department of Electrical Engineering 
Kyoto University 
Kyoto, Japan 
ABSTRACT 
PLATON (Programming LAnguage for Tree OperatioN) 
facilities of pattern matching and flexible backtracking, 
language is developed t~ simplify writing analysis programs 
The pattern matching process has the facility to extract sub- 
input sentence and invoke semantic and contextual checking fo?. 
actions between syntactic and other components are easily obt 
processing results in a failure, a message which expresses t: 
failure will be sent up. The control will be modified accoru 
enables us to write fairly complicated non-deterministic progr 
manner. An example of structural analysis using PLATON is als 
- ,le 
1. The 
1 language 
Erom thc 
Intc 
- 
-. rf 
Ise of . 
Th i :- 
;in:; 
ed " 
I Introduction 
In this paper we describe a new programming language which is designed 
to facilitate the writing of natural language grammars. 
A simple structural 
analysis program using this language is given as an example. There are two 
key issues in analyzing natural language by computer: 1) how to represent 
knowledge (semantics, pragmatics) and the state (context) of the world, and 
2) 
how to advance the programming technology appropriate for syntactic- 
semantic, syntactic-contextual interface. The point in designing a programming 
language is to make this kind of programming less painful. 
Tradi tima1 systems whi.ch represent grammars as a set of rewriting 
rules ustially 'have poor control mechanisms, and flexible interaction between 
thc syntact LC and other con~ponents is not possible. Systems in which rules 
of gr-;lnmlilrs .rlrc cn11,i-dilcd in proccdrlrcs, an. tllc other hand, make it possiblc 
to int t-rmis ttrct 6 yrltnc t 'lc and seniant ic anaIyst?s in an i r~tiinnte way. Llowcvcr, 
thest. systcos are apt. t.0 c.lestroy the intc.lligih:il ity and regularity of 
natural language gramars, because in these systems both rules and their 
control mechanisms are contained in the same program. 
Recently various syscems for natural language analysis have been 
developed T. Winograd's (1971) "PROGRAMMAR" is a typical example of 
procedure oriented systems. In this system the syntactic and other components 
can interact closely in the course of analyzing sentences. However, details 
of the program are lost in the richness of this interaction. LINGOL, 
developed by V. Pratt (1973) at MIT, is a language appropriate to syntax- 
semantics interface and in which it is easy to write grammars in the form of 
rewriting rules. The TAUM group at Montreal University (1971) has evolved a 
programming languagc name :I System-Q in which expressions of trees, strings 
and lists of them can be matched against partial expressions (structural 
patterns) containing variables and can be transformed in any arbitrary 
fast1 i 011. 
Thc augmcn tcd t ri111~ i L I on ilc twork (A1l'N) proposccl by W Woods (19 70) 
from our point of view gives an especially good framework for natural 
Xa~~guage analysis systems. One of- the most attractive features is the clear 
dtscrimination between grammatical rules and the control mechanism. This 
enables us to develop the model by adding various facilities to its control 
mechanism. 
The ATN model has the following additional merits :. 
30 
1. It provides power of expression equivalent to transformational 
grammars 
2. It mai11tai.n~ much of the rendability of context-free grammars. 
3. Rules of n grammar can be changed easily, so we can improve them 
through a t rial-and-error process whi1.e writing the grammar. 
4. It is possible to impose various types of semantic and pragmatic 
conditions on the branches between states. By doing this, close interactions 
between the syntactic and other components can be easily accomplished. 
However ATN has the following shortcomings, especially when we apply 
it to the parsing of Japanese sentences: 
1. It scans words one-by-one from the leftmost end of an input 
sentence, checks the applicability of a rule, and makes the transition from 
one state to another. This method may be well suited for English sentences, 
but because the order of words and phrases in Japanese sentences is relatively 
free, it is preferable to check the applicability of a rule by a flexible 
pattern-matching method. In addition, without a patte rn-matching mechanism, 
a single rewriting rule of an ordinary grammar is often to be expressed by 
several rules belonging to different states in Woods ATN parser. 
2. An ATN model essentially performs a kind of top-down analysls of 
sentences. Therefore recovery f ronl failures in predictf.on is most d1.f ficult. 
Considering these factors, we developed PLATON (a Programming 
Lhnguage for Tree-Ol>cratioN), which is based on the A1'N model and has various 
additional capabilities such as pattern-matching, flexible backtracking, and 
so on. As in SystewQ and LINGOL, PLATON's pattern-matching facility makes 
it easy to write rewriting rules. Moreover, it extracts substructures from 
the inputs and invokes appropriate semantic and contextual checking functions. 
31 
These may be arbitrary LISP functions defined by the user, the arguments of 
whi~n are the extracted substructures. 
A backtracking mechanism is also necessary for langu~e understanding 
as in other fields of artificial intelligence. During the analysis, various 
sorts of heuristic information should be utilizable. At any stage, analysis 
based on criteria which may relate to syntactic, semantic or contextual 
considerat ions taken separately may be unreliable. The result which fulfils 
all the criteria, however, will be a correct one. The program should be 
designed such that it can choose the most satisfactory rule from many 
candidates according to the criteria at hand. In further processing, if the 
choice is found to be wrong by other criteria, the program must be able to 
backtrack to the point at which the .relevant decision was made. In PLATON 
we can easily set up arbitrary numbers of decision points in the program. 
Then, if subsequent processing results in some failure, control will come 
back to the points relevant to the cause of the failure. 
11. Pattern-matching 
Before proceeding to the detailed description of PLATON, we will 
explain the representation schema for input sentences and parsed trees. The 
process of analyzing a sentence, roughly speaking, may be regarded as the 
process of transforming an ordered list of words to a tree structure, which 
shows explicitly the interrelationships of each word in the input sentence. 
During the process, trees which correspond to the parts already analyzed, and 
1hts which have not been processed yet, may coexist together in a single 
structure. We therefore wish to represent such a coexisting structure of 
trees and lists. A list structure is a structure in which the order of element 
is not changeable. On the other hand, a tree structure consists of a single 
root node and several nodes which are tied to the root node by distinguishable 
relations. Because relations between the root and the other nodes are 
explicitly specified, the order of nodes in a tree is changeable except for 
the root node which is placed in the leftmost position. Different matching 
schemas will be applied to trees and lists. 
The formal definition of such coexisting structures is as follows. 
<structure> is the fundamental data-structure into which all data processed 
by PLATON must be transformed. Hereafter we refer to this as the "structure" 
The formal definition of <structure) is: 
(list) ::= ($ ts tructures> ) 
(structures> : := 1 <structure> ( structures > 
<tree> : := ( node, I ( (node) < branches) ) 
<branches> : := (brancl~) 1 (branch, branches ) 
(branch) ::= (<relation> (tree, ) 
( node, : := Clis t > I !UBITRARY LISP-ATOM 
< relation > : := ARBITRARY LISP-ATOM 
A simple example is shown in Figure 1. 
Coexisting Structure of 
Trees and Lists 
Corresponding Expression 
in PLATON 
Figure 1 Expression of Structure in PLATON 
Two lists which have the same elements but different orderings (for example, 
(X A B C) and ($ A C B)) , should be regarded as different structures. On 
the other hand, two tree structures such as ( A ( R1 B ) ( R2 C )) and 
( A ( R2 C )( R1 B )) are regarded as identical. Besides the usual rewrite 
rules which treat such strings, s tructural patterns which contain variable 
expressions are permitted in PLATON. The PLATON-interpreter matches 
structural patterns containing variable expressions against the structure 
under process and checks whether the specified pattern is found in it. At 
the same time, the variables in the pattern are bound to the corresponding 
subs tructurzs . 
Variables in patterns are indicated as :X (X is an arbitrary LISP 
atom). The following can be expressed by variables in the above definition 
(1) arbitrary numbers of < structures> , that is to say, list elements in 
the definition of < list > (Figure 2, Ex. 1). We can also specify the 
Structural Pat terns Structures 
Results of Matching 
Example 1 
SUCCESS 
In A :K 
C C 
Example 2 
SUCCESS 
SUCCESS 
Figure 2 Illustration of Matching 
34 
number of list elements by indicating variables as :X+number. For 
example, the variable :D2 will match with two elements in a list. 
(2) arbitrary numbers of <branches > , in the definition of (tree > 
(Figure 2, Ex. 2). 
(3) < tree) in the definition of < branch) (Figure 2, Ex. 3). 
We shall fail such structural patterns (structure-1 > . 
By using the same 
variable several times in a pattern, we can express a structure in which the 
same sub-structure appears in two or more different places. 
The character 
! in a list indicates that the next element following the character is 
optional. 
111 Basic Operations of PLATON 
A grammar, whether generative or analytical. is represented as a 
directed graph with labeled states and branches. There is one state dis tin- 
guished as the Start State and a set of states called Final States. Each 
branch is a rewriting rule and has the following elements: 
(1) applicability conditions of the rule, typically represented as 
a structural pattern 
(2) 
actions which must be executed, if the rule is applicable 
(3) 
a structural pattern into which the input structure should be 
transformed. 
Each state has several branches ordered according to the preference of tho 
rules. 
When the control jumps to a state, it checks the rules associated 
with the state one-by-one until it finds an applicable rule. 
If such a rule 
is found, the input structure is transformed into another structure specified 
by the rule 
and the control makes the state transition. 
35 
In addition to the above basic mechanism the system is provided with 
push-down and pop-up operations. The push-down operation is such that in the 
process of applying a rule, several substructures are extracted from the 
whole structure by variable binding mechanisms of pattern-matching. Then each 
is analyzed from a different state. The pop-up operation is such that after 
each substructure is analyzed appropriately, control comes back to the 
suspended rule and execution continues. Usills these operations, embedded 
structures can be handled easily (See Figure 3) . 
Fipure 3 State Diapram 
Table 1 shows the formal definition cf a grarmr of PLATON (See follow- 
ing page). It shows that branches or rewriting rules in an ATN parser 
correspond to six-tuples (i e., (pcon) , <strx> , dcon> , ( c transr ) , 
( 4acts> ) , dendb . <strx> corresponds to the left side of a 
rewriting rule and describes the structural pattern to which a rule is 
applicable. ( strx > is, by definition 
(1) / or 
(2) s truc ture-1 
TABLE 1 Formal Definition of Grammar in PLATON 
(states, :: = <state> 1 <state> 4states> 
4 rules) . . - crule> <rules> 
-= - I 
4 rule > .. . . = ( ( pcon) cstrx, <con> ( <trans> ( 4 acts? tend 3 ) 
< trans> .. . . = I < transi0C trans, 
Iregister-name, 
< transit, : : = (( (state-name , <structure-2> ) derrorps) 
4 variable-nam- 
4 pros, 
.. - 
. . - <pro> I <pro> 4 pros> 
4 pro > 
.. = 
.. (EXEC <trans> ) I(TRANS ( (state-name> 4stsy> )) 
(end '> 
.. = (NEXT estate-name) <stry> 
I (NEXTB 4s ta te-name> rstry 2 ) 
((POP 4 stry> ) I (FM < failure-message ) 
.facts> 
.. . . = I 4 act><acts> 
<set > : : = <form , I (SR (register-name> d form> 
I (SU 4 regis ter-name > < form > ) 
~(SD <register-name > cf orm. ) 
< strx > 
.. = 
. . <structure-I> 1 / 
<s try, 
.. = 
. . <structure-2, 1 1 
dpcoTl),<con> : : = <form> 
( form) 
: : = (GR .<register-name> ) I (GV (variable- name > ? 
~(TR  structure-2. ) I(TR /) 1 ARBITRARY LISP FORM 
<variable- :: = :X (X is an arbitrary LISP atom) 
name 7 
<register- : : = /X (X is an arbitrary LISP atom) 
name) 
shows that a rule is applicable no matter what the structure under process 
is The variables used in (structure-1) are bound to corresponding 
substructures when matching succeeds. The results of Example 1 (See Figure 2) 
indicate that the variable :K is bound to the substructure (* ( B ( R1 C)) D ) 
The scope of variable binding is limited to within the realm of the 
particular rule. The same variable name in different rules has different 
interpretations. In this sense, :X-type variables in < s tructure-I> are 
called Local Variables. On the other hand, w2 can store certain kinds of re- 
sults from the application of rules in registers and refer back to them in 
different rules. These cons tftue variables which we call registers. They 
are represented by the symbols /X (X is an arbitrary LISP atom). 
Besides the pattern-matching, < peon> and < con ) can .~lso check 
the applicability of a rule. Certain parts of the results from the applica- 
tion of previous rules are contained in registers, not in the structure. 
We can check the contents of these registers by using < pcon > -part functions 
like GR, GU, etc. (these functions are listed in Table 2) and other LISP 
functions defined by the usual LISP function, DEFINE. (See following page for 
Table 2.) 
Semantic and contextual co-ordinatian between substruetures can be 
checked by using appropriate functions in the (con >-part of a rule. 
Semantic and contextual analyses cannot be expressed in the form of simple 
rewriting rules. These analyses have differing requirements such as lexical 
information about words which may in turn represent knowledge of the world 
and contextual information which may. express the state of the world. We can 
use arbitrary LISP-forms in the 4 con 7 -part, according to what semantic and 
contextual models we choose. 
TABLE 2 Functions of PLATON 
w 
Func t ion 
SR 
SV 
GR 
GV 
Argument 
(regis ter-name> 
LISP - < form> 
4 var iab le-nam- 
LISP - 4 form> 
(regis ter-name> 
(variab le-name> 
Effect 
SR stores the result of the 
evaluation of the 2nd argu- 
ment in the register. 
SV stores the result of the 
evaluation of the 2nd argu- 
ment in the variable 
GR get the content of the 
register 
GV gets the value of the 
variable 
TR 
SU 
SD 
GU 
PUSHR 
4 
Value 
.- - 
the result of the 
evaluation of the 
2nd argument 
A 
the result of the 
evaluation of the 
2nd argument 
the content of 
the register 
the value of 
the variable 
I <structure-2, 
or / 
, 
rregis ter-name, 
LISP - 4 form) 
Cregis ter-name 
LISP - <form> 
(register-name> 
4 regis ter-name > 
LISP - 4 form) 
, 
TR transforms the varizbles 
and registers in the struc- 
tural pattern into their 
values. 
SU sets the reigster of the 
- higher level processing 
SD sets the register of the 
the transformed 
structure 
the result of the 
evaluation of the 
2nd argument 
I 
lower level processing. 
GU gets the content of the 
register of the higher 
level. 
PUSHR is defined as the 
following. 
(SR r regis ter-name> 
(CONS 4 form, 
(GR <register- 
name3 1 1) 
the result of the 
evaluation of the 
2nd argument 
the content of 
the register 
the result of the 
evaluation of the 
2nd argument 
For example, suppose 
strx = ( ADJ ( TOK :N ))( N(TOK :N1 )) :I ) 
con = ( SEM :N :N1 ) 
Here TOK is the link between a word and its part of speech. :N and :N1 are 
the words of an input sentence. SEM is a function defined by the user which 
checks the semantic co-ordination between the adjective :N and the noun :N1. 
By this function SEM, we can search, if necessary, through both lexical 
entries and the contextual data bases. 
Uith this approach, if a certain syntactic pattern is found in the 
input structure, it is possible for an appropriate semantic runction to be 
called. Hence the intimate iiitcrcctions between syntactic and semantic 
components can be obtained easily without destroying the clarity of natural 
language grammars. 
Arbitrary LISP-forms can be also used in <act> -portion. They will 
be evaluated when the rule is applied. If necessary, we can set intermediate 
results into registers and variables by using the functions listed in Table 2 
(end > comprises f~ur varieties, and rules are divided into four 
types according to their ( end > types. 
1. NEXT-type : The < end+ is in the form (NEXT dstate-name, 4 s try> ) . 
The bstry) corresponds to the right side of a rewriting rule, and 
represents the transformed structure. A rule of this type causes 
state-transition to the dstate-name) , when it is applied. 
2. NEXTB-type: This rule also causes state-transition. Unlike with the 
NEXT-type, state-saving is done and if further processing results in 
some failures, control comes back to the state where this rule is applied. 
The environments, that is, the contents of various registers will be 
restored, and the next rule belonging to this state will be tried 
40 
3. POP-type : The (end > -part of this type is in the form (POP <s try > ) 
When it is applied, the processing of this level is ended and the 
control returns to the higher level with the value stry > . 
4. FM-type: The <end )-part of this type is in the form (FM <failure- 
message) ). The side effects of  he processing at this level, that is, 
register settings and so on, are cancelled (see section 4). 
In <stry > we can use two kinds of variables, that is, the variables used 
in dstrx > and registers. We find this structural pattern, called 
4s tructure-2 > , more suitable for writing transformational rules than 
Woods BUILDkoperation. By way of illustration consider the following: 
input string = CDE(A(Rl(* B)))FG) 
strx = (8 :I ( A ( Rl :N )) :J ) 
StrgT 
= (* (A ( FU (* :I :N ))( R2 /REG )) :J) 
the content of /REG = (G ( R3 H )) 
As the result of matching, the variables :I, :N and :J are bouad to the 
substructures (& C D E ), ($ B) and (3 F G ) respectively. The result of 
evaluating the < stry > is 
( ( A ( R1 (# C D E B ))( R2 ( G ( R3 H )))) F G ). 
If the rule is a POP-type one, then this structure will be returned to the 
higher level processing. If it is NEXT- or NEXTB-type, then the control will 
move to the specified state with this structure. 
IV Push-down and Pop-up Operations 
By means of NEXTB-type rules, we can set up decision points in a 
program. We can also do this by using push-down and pop-up operations. 
A 
rule in PLATON finds particular syntactic clues by its structural description 
( strx> ; and at the same time, extracts substructures from the input 
st'king From the structural description it is predicted that the substructures 
may have particular constructions, that is, they may comprise noun phrases, 
relative clauses or whatever. It is-necessary to transfer the subst~r~ctures 
to states appropriate for analyzing these constructions predicted and to return 
the analyzed structures back into the appropriate places In PLATON, these 
operations can be described in the Ctrans > -part of a rule. For example, 
suppose the <trans >-part of a rule is 
( (( Sl :K :K )) (( S2 ( :I :J)/REG )) ) 
When the cantrol interprets this statement, the substructures corresponding 
to the varuble :K and ( :I : J ) are transferred to the states S1 and S2 
respectively If the processings starting from these states are normally 
completed (by a POP-type rule), then the results are stored i? the variable 
:K and the regzster /REG. In this manner, by means of the push-down and 
pop-up mechanisms, substructures can be analyzed from appropriate states. 
Processing from kese states, however, may sometimes result in failure, 
That is, predictions that certain relationships will be found among the 
elements of substructures may not be fulfilled. In such instances the pushed 
down state will send an error-message appropriate to the cause of the failure 
by an Wtype rule. An FM-type rule points out that a certain error has 
occurred in the processing. If NEXTB-type rules were used in the previous 
processing at this level, control will go back to the most recently used 
NEXTB-type rule. If NEXTB-type rules were not used at this processing level, 
tne error-message specified by the FM-type rule will be sent to the <trans> 
part of the rule which directed this push-down operation (see Figure 4) 
Higher level 
Lower level p 
processing 
------------- 
/ 
/ \ 
/ \ 
I 
\ 
/ 
\ 
rocessing~ \ 
---I - 
I 
\ 
I 
\ i 
\ 
/ NExTB-type rules were 
not applied in this level 
Lower level processing 
c---- 
---- 
/This rule will be applied next. 
Figure 4 Illustration of Backtracking 
43 
According to these error-messages, 
control-flow can be changed 
appropriately. For example, we can direct processings by describing the 
4 trans > -part in the following way. 
( (( sl :K :K )( ERR1 ( EXEC (( S5 
:K :K 1) (( 56 ( 
:I :J ) /REG )))) 
( ERR2 (TUNS ( S8 /))I 
((S2 (j6 :I :J ) /REG )) ) 
In the above example, 
the processing of the substructure :K from the state 
S1 will produce one of the following three results. 
According to the 
returned value, the appropriate step will be taken: 
(1) 
Normal return: 
the processing of :K is ended by a POP-type 
rule. 
The result is stored in the variable :K and the next push-down 
performed, that is :I : J ) will be transferred to the state S2. 
(2) Return with an error-message: the processing of :K results in 
a failure and an FM-type rule sends up an error-message. 
If the message is 
ERR1, them :K and :f :J ) will be analyzed from the states S5 and S6 
respectively (EXEC-type). 
If it is ERRS, the interpreter will give up the 
application of the present rule, and pass the control to another state S8 
(TRANS- type). If it is neither ERRl nor ERR2, the same step as (3) will be 
taken 
(3) Re turn with the value NIL: the processing Prom the state 
S1 
will send up the value NIL if it runs into a blind alley, that is, there are no 
applicable rules. The interpreter will give up the application of the present 
rule and proceed to the next rule attached to this state. 
Mechanisms, such that control flow can be appropriately changed 
according to the error-messages from lower level processings are not found in 
Woods ATN parser, We can obtain flexible backtracking facilities by combining 
these mechanisms with NEXTB- type rules. 
V A Simple Example 
We are now developing a deductive question-answering sys tern with 
natural language inputs -- Japanese sentences. The internal data-base is 
assumed to be a set of deep case structures of input sentences. We adopted 
and modified Fillmore's (1968) case grammar to analyze the input of Japanese 
sentences. Japanese is a typical example ofan SOV-language in which the 
object and other constituents governed by a verb usually appear before :he 
verb in s sentence. A typical construction of a Japanese sentence is shown 
in Figure 5. 
Figure 5 Typical Construction of a 
Japanese Sentence 
A verb may govern several noun phrases preceding it. A relative clause 
modifying a noun may appear in the form -- verD + noun -- The right 
boundary of the clause is easily identified by finding the verb. The left 
boundary is often much more difficult to identify. In Figure 5, the noun 
phrase NPkr is a case element of the verb V On the other hand, the noun 
phrase NPi is governed by the verb Va Because the rule of projections holds 
in JaFnese as in other languages, all the noun phrases between NP c'+ I and 
V are governed by V 
, and the noun phrases before NP,: are governed by VA 
However, in the cou.rse of analysis, such boundaries cannot be determined 
uniquely. The analysis program fixes a temporary boundary and proceeds to 
the next step in processing If the temporary boundary is not correct, the 
succeeding processing will fail and the control wil come back to the point 
at which the temporary boundary was fixed. 
Now we will show a simple example of structural analysis by PLATON 
The example explains how the backtracking facility is used in analyzing 
Japanese sentences. Because we want to visualize the operations of PLATON 
without bothering with microscopic details of Japanese sentences, we will 
take an imaginary problem as an example. 
The parser which is written in 
PLATON is described in another paper by M Nagao and J. Tsujii (1976) 
An input string is assumed to be a list. The elements of the list 
are integers and trees are in the form of ( X (SUM 0)). Here 'X' may be 
regarded as a term modified by 'SUM 0' These two kinds of elements are 
arranged in an arbitrary order, except that the last element is the tree 
(X(SUM 0)). The following is an example of an input string: 
( 5 2 13 (x (Srn 0)) 3 1 (X (SUM 0)) 2 2 ( x (SUM 0)) ) 
Figure 6 An Example String to be Analyzed 
The result of the transformation is expected to be in the following fotm: 
( (X (Sm4)) (XSm6)) (X (Sm9)) 1 
This result is regarded as representing the following relationships between 
integers and 'X' . 
The number associated with an 'X' by the relation 'SUM' shows the sum of 
the integers which are governed by the X. We can look upon the relations 
r 
between integers and an X' as the relations between noun phrases and the verb 
in Japanese sentences. 
The result of the analysis is assumed to satisfy 
the following conditions. 
46 
(1) Governor-governed relationships between integers and an 'X' must obey 
the projection rule (i.e., clauses do not overlap). 
(2) As a simulation of a semantic restriction, we attach a condition that 
the sum of the integers governed by an 'X' should not erceed ten. 
(3) As a simulation of a contextual restriction, we attach the condition 
that a result (F (X (SUM num-1)) (X (SUM num-2)) . . . . (X (SUM num-N)) ) 
should maintain the relation, numl 4 num-2 - 5 . . . .. <um-N. 
- 
set 05 rules is shown in the following. The corresponding state- 
diagram is shown in Figure 7. 
START 
NEXTB 
POP 
NEXT 
Fipure 7 State Mapram of a Si wple Example 
SL? -1- strx: = :I :I1 (X (SUM :N)) :J) 
con: = ( GREATERP 10 (PLUS : N : 11) ) 
act: = ( (SV :N (PLUS :N :I1 )) 
(PUSHR /REG :Ill ) 
end: = (NEXT SUMLm C)c :I (X (SUM :N)) : J) ) 
-2- strx : = (* :I (X (SUM :N)) rJ) 
con: = (CONTEXTCHECK /RESULT (TR (X (SUM :N)))) 
act: = NIL 
end: = (NEXT BACKTRACK /) 
-3- strx: = (% :I (X (SLW :N)) :J) 
con: =: T 
act: = NIL 
end: = (FM-ERROR) 
-4- strx: = (8 ) 
con: = T 
act: = ( (SR /RESULT (CONS 'X /RESULT )) ) 
end: = (POP /RESULT) 
BACKTRACK 
-1- strx: = :I (X (SLW :PI)) :.J) 
con: = T 
act: = ( (SR /REG NIL) 
(SR /RESULT (APPEND /RESULT ( TR (X (SUM :N))))) ) 
end: = (NEXTB SUMUP (S# :I :J )) 
-2- strx: = (* :I (X (SUM :N)) :J) 
con: = T 
act: = ( (POPR /TEMP /REG) 
(SV :N (MINUS :N /TEMP)) ) 
end: = (NEXT BACKTRACK 01: : I /TEMP (X (SUM :N)) : J) ) 
The input string is the! list shown in Figure 6. Since the start state 
is SWP, the first rule attached to this state is applied. This rule will 
find the leftmost 'x' and an integer just before the 'x' (by SUMUP -I-, strx). 
The variable :I1 is bound to tius integer. This integer is added to the sum 
of the integers, :N, if the total does not exceed ten (SUMUP -I-, con). 
PUSHR, used in the < act > -part, is a PLATON function which puts the 
second argument on the head of the first argument (SUMUP -I-, act) After 
this rule is applied, the control will enter the state SUMUP again (SW -I-, 
end). That is, this rtile is applied until there are no ~ntegers before the 
first ' X' or the sum of the integers. exceeds ten. As the result, the 
environment is the following: 
structure under processing 
= qY 5 (X (SUM 6)) 3 1 (X (SUM 0)) 2 2 (X (SUM 0)) ) 
relationship temporarily fixed be tween integers and 'X' 
content of /REG 
=(213) 
The second rule of SUMUP will be applied next. This rule checks by its 
1 con> part whether the result at hand satisfies the third condition, 
that is, the contextual restriction. Because the content of /RESULT is NIL, 
the function CONTEXTCHECK returns the value T (SllMUl' -2-, con). So this 
rule is applicable. Control makes the state-transition to the state 
BACKTRACK (SUMUP -2-, end.) Because the first rule of BACKTRACK is a NEXTB- 
type yule, state-saving is performed. 
That is, tbe following environment is 
saved: content of /REG = ( 2 1 3) 
content of /RESULT = NIL 
structure under processing = 
# 5 (X (SUM 6)) 3 1 ( X (SUM 0)) 2 2 (X (SUM 0 )) ) 
49 
By this rules, the registers /REG and /RESULT are set as follows ( BACKTRACK 
-I-, act). /REG : = NIL 
/RESULT: = ( ( X (SUM 6 ) ) ) 
And the structure is transformed to 
(X 5 3 1 (X (SUM 0)) 2 2 (X (SUM 0)) ) 
A XEXTB- type rule causes a state transition as does a NEXT-type rule. 
So 
control returns to the state SUMUP (BACKTRACK -I-, end) . At tt.is state, a 
process similar to the one described above is performed. As a result, the 
following governor-dependent relationships are es tablished. 
Here the bold lines indicate the newly established relationships. By the 
first rule of BACKTRACK the following environment is saved. 
content of /REG = ( 5 3 1 ) 
conten& of /RESULT = ( ( X (SLW 6 )) ) 
structure under processing = (# 
(X (SUM 9)) 2 2 (X (SUM 0)) ) 
And /REG and /RESULT are set as the following (BACKTRACK -1-, act) . 
/REG: = NIL 
/RESULT: = ( (X (SUM 6)) (X (SUM 9)) ) 
The transformed structure is (BACKTRACK -10, end) 
Cjl 2 2 (x (SUMO)) 1 
The control is transferred to the state SUMUP. By applying the first rule 
of this state repeatedly on the above structure the following structure is 
obtained. 
(j( 
( x SUM 4)) 1 
50 
However this result does not satisfy the con textual restriction. 
So the application of the second rule of SLJMl.JP fails because the function 
CONTEXTCHECK used in <con ) -part returns the value NIL (SUMUP -2-, con) 
That is : 
con textcheck [( (X (SUM 6) ) (X (SUM 9) ) ) : (X (SUM 4)) 1 = NIL 
The third rule, therefore, will be applied next. Because this rule is a 
%type rule (SUMUP -3-, end), it causes an error and control comes back to 
the point at which a NEXTB-type rule was applied most recently. The saved 
nviro~ment is restored. This is: 
/REG: = ( 5 3 1 ) 
/RESULT: = ( (X (SUM 6)) ) 
structure under processing: = (* 
(X (SUM 9)) 2 2 (X (SUM 0)) ) 
Then by applying the second rule of BACKTRACK, the governor-governed 
relationship established lastly in the previous process is cancelled. The 
structure and the register /REG are changed as below (BACKTRACK -2-, act) : 
/REG: = ( 3 1 ) 
structure under processing: ( 5 (X (SUM 4)) 2 2 (X (SUM 0 )) ) 
Control enters the BACKTRACK state again. The application of the 
first rule saves the environ~wnt : 
content of /REG = ( 3 1 ) 
content of /RESULT = ( (X (SUM 6)) 
structure under processing = (1< 5 ( X (SUM 4)) 2 2 (X (SUM 0)) ) 
That is, the relationship indicated by the dotted line in the following is 
cancelled: 
Control transits to the state SUMUP (BACKTRACK -I-, end) and a 
similar process is performed. However, because the governor-governed 
relationship between the integer 5 and the second 'Xr is cancelled, the sum 
of the integers governed by the first 'X', ( 2 1 3 ), is greater than that 
of the second ' X' , ( 3 1 ) . The contextual condition, therefore, is not 
fulfilled, and the application of the second rule of SUMUP will not succeed. 
So the temporarily established relationships will be cancelled one-by-one as 
follows. 
After these relationships have been cancelled, the desired result is obtained 
by the following sequence. 
52 
At the final stage of the processing, the fourth rule of SW a 
POP-type rule, is applied and returns the value 
( (X (SUM 4)) CX (SUM 6)) (X (SUM 9) 1 
VI Conclusion 
We have described a programming language called PLATON for natural 
language processing. The language has several additional capabilities beyond 
the ATN parser of W. Woods. 
Grammars written in the language not only maintain clarity of 
representation but also provide adequately a natural interface between the 
syntactic component and other components. By means of the pattern-matching 
facility, we can write grammars in a quite natural manner. And because of 
the PLATON variable binding mechanism, semantic and contextual LISP fuactions 
a easily incorporated in syntactic patterns. 
Flexible backtracking mechanisms and push-down operations make com- , 
r, Acatecr non-deterministic processing possible in a very simple way. 
We are now deveLoping an analysis program for Japanese using this 
language. The program can accept fairly complicated sentences in a textbook 
of elementary chemistry It can utilize the lexical and contextual information 
of chemistry adequately during the analysis. Such information in our system 
is expressed in the form of a semantic network similar to that of R. F. Simmons 
(1973). 
Perhaps, PLATON itself must be equipped with more semantics and 
context-oriented operations such as specified lexical descriptions and functions 
using them. 
However, what description method is most efficient, and moreover, 
what semantic information must be stored in the lexicon, are not yet entirely 
clear. So, as the first step, PLATON leaves many parts of these problems for 
53 
the user to specify by LISP programs. 
PLAMN is written in LIS1'1.5 and 
implemented on a FACOM 230-60 at the Kyoto University computing center and a 
TOSBAC-40 mini-computer in our laboratory. The interpreter of PLATON itself 
requires only 4.5 K cells. 
American Joura id of (Clompu~ationaP 1 inguisties Micicrofiche 37 : 54 
IEMBERSHIP STATERENT 
................ Members paid through 1976 5 
................ Members paid through 1975 571 
................ Members paid through 1974 121 
............. Exchange and gift subscriptions 47 
American Journal of Computational Linguistics 
Microfiche 37 : 55 
EDITOR'S REPORT 
During 1975, this Journal published 20 contributions through 
the ordinary channels of review; it also published two micro- 
fiches of abstracts from the Varna conference and five 
microfiches of proceedings of the 1975 Annual Meeting of the 
Association for Computational Linguistics. 
A directory of 
members also appeared. 
Early in 1975, the Journal announced an experimental procedure 
for review of submiesions; the author could submit a 600-word 
summary for distribution to all members of the Editorial 
Board, €If the last 30 submissions, 15 have used this proce- 
dure. Contributors and Board members appear satisfied with 
the results. 
The number of microfiches distributed in the 1975 subcription 
was 22, up from the 14 distributed to I974 members. 
But the 
number of frames increased from 900 to 1700, roughly; as 
anticipated, the average number of frames filled is going up 
Since the first announcement of the Journal, 86 submissions 
have been received. Their present status is shown at the top 
of the following frame. 
EDITOR S REPDRT 
31 have been published 
18 are with the author for ninor revision 
15 are with the author for major revision 
16 have been rejected 
3 are in process 
2 have been returned as unsuitable or withdrawn 
At the end of March, 1975, 10 manuscripts were out for ninor 
revision; 4 of them have been published. Of the 7 then out 
for major re sion, two have been published. We might thus 
expect to publish as many as 7 or 8 of the manuscripts now 
out for revision, making a total of about 36 of 81 received; 
our ultimate acceptance rate may be around 45%. 
The financial support of the National Science Foundation, 
which continued through 1975, and the unpaid help of both 
the State University of New York and the Xerox Palo Alto 
Research Center (the latter is processing bibliographies and 
indexes) have been most significant in establishing the new 
Journal. The work of the Editorial Board, editorial aides, 
and the staff of the Center for Applied Linguistics are also 
irnpor tant . 
American Journal of Computational Linguistics Microf. 
SURVEY OF MEMBERS 
PRELIMINARY REPORT 
The last mailing of the L975 subscription included a question- 
naire about the level and direction of members interests, 
their use of the Journal, and their preferences among three 
forms of publication: microfiche, paper, and mixed. 
By the end of January, 200 replies had arrived and the rate of 
arrival had diminished * The present report makes no interpre- 
tations or explanations; the figures are here for the Editorial 
Board, the Executive Committee, and the members to examine. 
All written comments received in response to the questionnaire 
are presented following the numerical data. 
Members who have strong opinions about the policy the Associa- 
tion should adopt may write to any officer of the Journal or 
Association; but prompt action is needed in order to permit 
planning and development of manuscripts for the 1977 volume. 
* The issue left the editor on November 7; 
the questionnaire 
was to be returned "by December 31!' 
Due to unforeseen delays, 
the issue reached readers beginning mediately after Christmas 
in the Washington area, and late in January in Hawail and abroad. 
RESPQNSES TO SINGLE QUESTIONS 
1 How many AJCL microfiches have you read? 
None 30 
One 20 
2 to 5 61 
6 or more 88 
No response 1. 
2. How can you read a microfiche? 
Own a microfiche reader 50 
Go to a convenient place 80 
Go to an inconvenient place 64 
No means available 6 
3. How many AJCL opaque cards have you examined carefully or 
read through? 
None 6 
1 to 5 18 
6 to 10 40 
11 or more 135 
No response 1 
4. How significant is the material in AJCL to you? 
Highly significant; I need much of the content of 
AJCL in my work 32 
Fairly significant; I sometimes need information 
of the kind I have seen published in AJCL 131 
Hardly significant; I rarely need anything from AJCL 32 
Not significant; nothing I have seen in AJCL would 
be useful to me 2 
No response 3 
SURVEY OF MEMBERS 
What do you want ACL to supply to you? 
Original papers 181 
Survey papers 161 
Bibliography 151 
News 149 
CHECK AS MANY 
AS APPROPRIATE 
6. Judging from your own opinion and what you know of your 
colleagues, does AJCL probably lose contributions to other 
journals because they are printed and AJCL is on micro- 
fiches? 
Yes, AJCL's format loses contributions 75 
No, AJCL's format probably does not lose contributions 82 
No response 43 
7. Is a microfiche printer available to you if you want 
full-size copy? 
Yes 86 
No 106 
No response 8 
8. If AJCL offered paper copies of the articles on microfiche, 
sent by mail at S2 per article (up to 97 pages), how many 
would you have bought in 1975? 
None 88 
One 21 
2 or 3 64 
4 or more 21 
No response 6 
Which of the following plans do you want AJCL to carry out? 
Microfiche publication; about 2000 manuscript pages 
per year 116 
Paper publication; about 500 manuscript pages per year 
without an increase in dues 24 
Mixed publication; about 250 manuscript pages per year 
printed on paper, supplemented by about 1000 manuscript 
pages on microfiches 44 
No response 16 
10. If AJCL continues to publish on microfiches, but offers a 
subscription on paper at a high price as an alternative, 
would you pay $20 to $30 extra for yourself? 
Yes 
No 
No response 6 
11. If AJCL continues to publish on microfiches, what do you 
expect to do? 
Resign from the As 
Pay dues without r 
Buy (or borrow) a 
Continue a satisfa 
No response 
sociation 
eading the 
reader 
ctory arra t for reading fiches 133 
12. 
Indicate your range of interests by checking as many boxes 
as appropriate. 
Leave blank the topics that do not interest 
you. 
(The purpose of this question is guidance for the 
Editorial Board.) 
The answers are tabulated on the next frame and displayed 
in a graph on the next following frame. 
Thereafter the 
cross - tabulation of groups of questions begins. 
Topical category 
Acoustic phonetics 
Acoustic recognition 
Acoustic synthesis 
Character recognition 
Text input systems 
Text storage systems 
Parsers 
Sentence generators 
Semantic interpretation 
Discourse analysis 
Semantic memory structures 
Mathematical linguistics 
Lexicography 
Lexicology 
Concordance systems 
Vocabulary statistics 
Inference programs 
Programming languages 
Information structures 
Indexing sys terns and theories 
Document classification 
Machine translation 
~sychological applications 
Social science applications 
Analysis of literature 
Instructional s ys terns (CAI) 
Level of interest 
Strong Slight 
29 40 
31 57 
31 42 
34 56 
79 43 
74 40 
105 54 
69 64 
131 33 
104 48 
120 30 
54 60 
43 45 
38 39 
21 64 
32 62 
82 38 
87 55 
100 46 
58 53 
38 46 
66 75 
58 52 
38 42 
35 51 
45 62 
0 
eye#+ input systems 
vo Le nitolog 
" 6ericyro~~L~ 
Y ? .D~UM.R+ el~sitic*+ioh 
social ~io*e~‘a~~liur;&s 4 C harobtet NW~ n; tien 
i 
30 Acoust;r syfirle~is+o a .Voab~lu stdistics 
P kArous+i c r eogm;+ion 
Acouttie pho~g+ic5 
I 
20 BwIpis ~iteratsrred o CWWO~O~~MC~ s stems 
Y 
RESPONSES BY LEVEL OF SIGNIFICANCE OF AJCL 
In these tables, each cell reports the number of members (out 
of 200) who give specified answers to question 4 (How signifi- 
cant is the material in AJCL to you?) and another question, 
jointly. In the first table, the number of members who find 
AJCL fairly significant and have read 6 or more microfiches 
is 48. 
1. How many microfiches have you read? 
Significance Number of microfiches read 
0 1 2-5 6+ NA 
High 2 4 5 20 1 
Fair 19 10 43 59 
 LOW;^ 9 6 13 9 
"Includes Hardly or Not significant and no response. 
2. How can you read a microfiche? 
Significance 
High 
Fair 
Low 
Method of reading microfiches 
Own Convenient Inconvenient No 
reader place place means 
5 14 12 1 
35 56 3 7 3 
10 10 15 2 
3. How many WCL opaque cards have you examined or read? 
Significance Number of cards read 
O 1-5 6-10 11+ NA 
High 2 4 25 1 
Fair 
5 6 29 91 
Low 
1 10 7 19 
5. What do you want AJCL to supply tc you? 
Significance 
High 
Fair 
Low 
Original Sur-vey dibliography News 
papers paper s 
31 29 27 30 
120 103 102 97 
30 29 22 22 
6. Does AJCL probably lose contributions? 
Significance Does lose Does not lose No response 
High 17 12 3 
Fair 44 58 29 
Low 14 
12 11 
7. Is a microfiche printer available to you? 
Significance Printer available 
Yes No NA 
High 13 16 3 
Fair 57 71 3 
Low 16 19 2 
8. How many articles would you buy from AJCL at $2? 
Significance Number of articles 
0 1 2-3 4+ NA 
High 13 2 10 6 1 
Fair 57 13 46 11 4 
Low 18 6 8 4 1 
9. Which plan do you want AJCL to carry out? 
Significance Microfiche Paper 
Mixed No response 
High 
19 1 8 4 
Fair 
77 14 30 10 
Low 20 
9 6 2 
10. If AJCL publishes microfiches, would you pay $20 for paper? 
Significance Yes No NA 
High 
2 29 1 
Fair 
8 119 4 
Low 3 33 1 
11. If AJCL continues on microfiches, what will you do? 
Significance Resign Pay dues, Buy a Continue NA 
not read reader as now 
High 
Fair 
Low 
RESPONSES BY CONVENIENCE OF READING MICROFICHES 
In these tables, each cell reports the number of members (out 
of 200) who give specified answers to question 2 (How can you 
read a microfiche?) and another question. 
1 How many microfiches have you read? 
Convenience 
High* 
Low$:* 
Number of microfiches read 
0 1 2-5 6+ NA 
5 8 46 70 1 
25 12 15 18 
* Own a reader or go to a convenient place 
**Go to an inconvenient place or have no means 
3. How many AJCL opaque cards have you examined or read? 
Convenience 
High 
Low 
Number of cards read 
O 1-5 6-10 II+ NA 
3 13 26 87 1 
3 5 14 48 
4. How significant is AJCL? (See table above.) 
5. What do you want AJCL to supply to you? 
Convenience 
High 
Low 
Originai Survey Bibliogrpahy News 
pr;l.pers papers 
6. Does AJCL probably lose contributions? 
Convenient e Does lose Does not lose No response 
High 32 73 25 
low 43 9 18 
7. Is a microfiche printer available to you? 
Convenience Printer available 
Yes No NA 
High 65 59 6 
Low 21 47 2 
8. How many articles would you buy from AJCL at $2? 
Conveni enc e Number of articles 
O 1 2-3 4+ MA 
High 69 16 36 8 L 
Low 19 5 28 13 5 
9. Which plan do you want AJCL to carry out? 
Convenience Microfiche Paper Mixed No response 
Hign 93 6 24 7 
Low 23 18 20 9 
10. If AJCL publishes microfiches, would you pay $20 for paper? 
Convenience Yes No NA 
High 3 125 2 
Low 10 56 6 
11. If AJCL continues on microfiches, what will you do? 
Convenience Resign Pay dues Buy a Continue NA 
not read reader as now 
High 1 5 11 112 1 
Low 9 13 22 21 5 
RESPONSES BY CONVENIENCE AND SIGNIFICANCE 
In the following tables, each cell reports the percentage of 
members who indicate a given degree of significance for the 
material in AJCL, a given degree of convenience in reading 
microfiches, and a specified answer to a third question. For 
example, 91 of 200 members find the material fairly signifi- 
cant and have a convenient means of reading microfiches. Of 
those 91, 48 have read 6 or more microfiches; 48 constitutes 
53% of 91, so the entry in the indicated cell is 53. 
1. Percentage who have read 6+ microfiches 
Significance Convenience 
High Low 
High 74 46 
Fair 53 27 
Low 40 6 
3. Percentage who have read 11+ cards 
Significance Convenience 
High Low 
High 84 69 
Fair 67 7 5 
Low 50 53 
5. Percentage who want AJCL to publish original articles 
Signific 
High 
Fair 
Low 
ance Convenience 
High Low 
95 100 
9i 92 
85 76 
Percentage who want AJCL to publish survey articles 
Significance Convenience 
High Low 
High 84 100 
Fair 81 72 
LOW 75 82 
Percentage who want AJCL to publish bibliography 
Significance Convenience 
High Low 
High 79 92 
Fair 78 77 
Low 55 65 
Percentage who want AJCL to publish news 
Significance Convenience 
High Low 
High 95 92 
Fair 71 80 
Low 55 65 
6. Percentage who think AJCL loses contributions 
All respondents Excluding no respoilse 
Significance Convenience Significance Convenience 
High Low High Low 
High 32 84 High 35 92 
Fair 25 52 Fair 33 78 
Low 15 65 Low 23 85 
7. Percentage who have a printer available 
High 
Fair 
Low 
Significance Convenience 
Hign Low 
47 30 
48 32 
60 24 
8. Estimate of number of printed papers that would be pur- 
chased, per member in the category; assuming 2.5 papers 
if the answer is 2 or 3, and 5 papers if the answer is 
4 or more. 
Significance Convenience 
High Low 
High 1.2 2.4 
Fair 1.3 1.7 
Low 0.5 2.2 
9. Percentage who prefer each publication policy 
Significance Convenience 
High 
Fair 
Low 
High Low 
Micro Paper Mixed Micro Paper Mixed 
63 5 21 54 31 
71 4 20 30 25 30 
80 5 10 24 47 24 
The percentages for the three policies should add to 100 
in each category of convenience and significance, but the 
percentage who did not answer is not shown. 
10. Percentage who would pay $20 to $30 extra for paper 
Significance Convenience 
High Low 
High 5 8 
Fair 2 15 
Low 0 18 
SURVEY OF MEMBERS 
11. Percentage who intend to take each action if MCL conti~ues 
to publish microfiches 
Convenience Significance Resign Not Buy a Continue 
read reader as now 
High High 0 0 11 89 
Fair C) 5 9 85 
Low 5 0 5 90 
Low High 0 8 46 31 
Fair 12 17 27 35 
Low 24 29 29 18 
Each row of the table should sum to 100%, but some did not 
respond to question 11. 
LONG COMMENTS 
Here's my survey questionnaire. Anent the decision as to where 
to go from here, I have a very strong recommendation to make. I 
think you're in a pioneering position on something very important, 
and should stick with it, certainly for at least two or three 
more years if at all possible. The major difficulty and source 
of objections, I think, is not any of the arguments against 
microfiche that you outline, but simply the difficulty people 
have in adopting radically new habits in such a fundamental ac- 
tivity as reading. 
Which Srings me to the second point. I think that as a pioneer, 
the Editorial Board of the AJCL has a responsibility to try to 
make the experiment- succeed, and not just sit back and watch. 
Specifically, I think it's very important to make available to 
subscribers a microfiche reader which works reasonably well, is 
not terribly expensive, and most importantly, is easy to acquire. 
I would bet that if you arranged with a manufacturer or dealer 
to supply a small, practical reader in the $25-$35 price range, 
which people could order simply by writing their name and address 
on an addressed, postage-paid return card included in an MCL 
packet, at least a hundred would be sold within a month You 
wbuld have done your readers a service, and would alsb have 
broadened the base of those who are in a position to make use of 
the micro£ iche format. 
There are a variety of details to-be considered in implementing 
such a plan, for instance, that. the supplier should be prepared 
to make exchanges for defective units, but I'm sure that the 
members of the Editorial Board are as alert to these as I am. I 
would appreciate hearing what reaction people have to this 
suggestion. 
Paul G. Chapin 
National Science Foundatioa 
Washington, D.C. 20550 
I am quite pleased at receiving the AJCL on microfiche and would 
hope that other organizations to which I belong would at least 
offer optional microfiche subscriptions to their journals. I 
find myself inundated with prfnted publications and am running 
out of space to store these valuable documents; I do not want my 
office or my home turned into a library. 
When I receive a copy of the AJCL, I read through the opaque 
cards looking for paper summaries that interest me. When I find 
such a paper, I use the department's microfiche reader to browse 
through it. Occasionally I have the library make a paper copy 
for me so that I can study papers without access to a microfiche 
viewer. 
One service which I think might be useful for ATCL subscribers 
without such convenient access to hard-copy microfiche machines 
would be an opaque card which one could use to order paper copies 
of selected articles. 
James R. Rhyne 
Department of Computer Science 
University of Houston, Texas 77004 
The microfiche journal is the best thing ACL has got going for 
it. In che last couple of years, I've been letting my member- 
ships in a number of organizations lapse. One of the principal 
reasons is that I simply can't cope with the flood of paper 
which results from memberships. My house and office get clut- 
tered up with stacks of magazines that can't be chucked because 
one article in a dozen looks as if it ought to be read some time. 
It's gotten to the point where new issues go straight from the 
mailbox to the recyclery. For this "servfce", one pays extra. 
Microfiche solves that; I've got two years' worth of ACL output 
in a 6" x 6%'' metal box, and there's room far two more years. 
If I ever run out of room for metal boxes, I can cull out se- 
lected cards. 
11 
By all means, make your experiment" permanent! 
Anonymous 
More notes on your questionnaire! 
(1) I have been getting IEEE Computer, IEEE Computer Transactions, 
and IEEE Spectrum since 1969 on microfiche. 
I no longer have 
paper copies of these earlier issues. 
I bought a cheap micro- 
fiche reader in 1969. 
(2) I bought a Realist (Xerox 320) l.fi .rears ago. 
(3) 1 also get now AJCL Journal 
(yours), Scientific American 
(Bell & Howell) , IEEE Software Engineering regularly on micro- 
fiche. 
(4) I buy selected titles from NTIS and University Microfj-Lms 
(Micromedia--kids books) . 1 subscribe to NTIS computer xeekly 
abstracts @$3O/yr. No company reimbursement. 
(5) After a years research I have produced my first amateur 
microfiche at home which dops on a Bruning. (So I now will be 
a publisher!) 
(6) I am a microfilm/microEiche COM expert in Management Systems 
at work (only part time). I am a systems analyst on data center 
management software . 
(7) Being an engineer and an analyst AJCL is currently a peri- 
pheral area of interest. I would not have joined your organi- 
zation if your journal was only on paper. 
I got into microfilming 
because I was running out of space. 
(8) I believe I will be able to keep a tremendously large number 
of journals, magazines, professional or pleasure reading books if 
they are on microfiche. By keeping good inventories in a safety 
deposit box I will have good insurance and replacement in case of 
fire. 
49) Able to microfilm (fiche) myself I will be able to archive 
(and back up against fire) any book or paper I work on. 
It would 
be easier to correspond and exchange papers with someone around 
the world or U.S. 
Lower postage. 
(10) I would like to see all hard bound and paper back books, 
all magazines available on microfiche. 
The savings for the 
Library of Congress and copyright off ice would be tremendous. 
These fiche should also be available 
to individuals from the 
publishers. 
(11) I also get the ACM Communications, and ACM Surveys on fiche 
I trust I will see the results of your survey, and maybe my let- 
ter, in the next AJCL Newsletter. 
Neil R. Karl 
21425 Duns Scotus 
Sauthfield, Michigan 48075 
Your recent questionnaire on the AJCL format invited additional 
comment I may be biased but I think micropublication of our 
journal is a superb answer to the twin problems of skyrocketing 
publication/distribution costs and the need to disseminate more 
information to students of this growing field. 
No serious student or scholar should be without a microfiche 
reader. Good compact readers from $100 (the Bell & Howell 
Brief CaS e reader) to $214 (the Washington Scientific Industries 
Mini-Cat) are available. Surely a modest investment for a 
life-time of use. 
I do think, however, that you can more effectively package the 
microfiche for increased user convenience. There are two basic 
approaches: 1) Bind fiche envelopes and the eye-legible cards 
together before they are distributed to subscribers or 2) provide 
subscribers with a conyenient microfiche binder file. 
Thomas F. Deahl, Microdoc 
8 I 5 Parnont~r T.ane Phi Iarlelnhia 19119 
" -HI r --- --.a ----- I - -------- r---- ----- 
It is my opinion that far from being a disadvantage, microfiche 
is a positive advantage. Already a large amount of my home and 
office space is devoted to storing technical journals which I 
have saved over the years, and as time passes the problem will 
only get worse. I only hope that you will continue to distribute 
AJCL on fiche, and that other journals will follow your example. 
I do, however,find the small size of the opaque cards to be in- 
convenient. Perhaps you could combine all the opaque cards in 
each issue into one or two 85x11" sheets, which could be folded 
when mailed, then spread open and collected in a looseleaf or 
other book when received. 
Norman Haas 
35-20 Leverich Street 
Jackson Heights, N .Y. 11372 
While having the freedom to publish very detailed accounts of 
large programs is useful I think that a more space conscious 
medium would encourage more concise descriptions. The freedom to 
be verbose does not strike me as a good idea in this field! 
Also, it is difficult to judge the effect of the microfiche format 
on soliciting manuscripts because the journal is new and thus has 
no developed prestige. 
Unsigned 
Although I do not object to microfiche format, I find the orga- 
nization of opaque cards and fiche sometimes confusing and easily 
disordered. 
Unsigned 
I'd like to see ACL offer--at a good price--a fiche reader. 
My intuitive requirements are 1) "book-page" size image, 2) stow- 
able, 3) transportable, 4) under $100. Comment? 
Peter Z. Ingerman 
40 Needlepoint Lane 
Willi~orq. N =J. -080.46 
I 12lieve the best alternative would be to 1) continue journal on 
micro£ iche , 2) arrange to sell inexpensive readers to subscribers 
through ACL (or widely publicize where and for how much readers 
are available) 
3) offer hard copy versions of articles by indi- 
vidual order. 
Unsigned 
I am a computer programer by profession and my interest in lin- 
guis tics extends only to those areas which are directly applica- 
ble to computers. Even there my interest is more concerned with 
the practical side rather than with esoteric theory. Since most 
of what you print can only be considered as future referent mate- 
rial by me, your printing of the journal on microfiche is eminently 
satisfying to me; in fact, if you switched entirely to paper, I 
would have to seriously consider my continued receiving of the 
j ournal . 
Robert D. Kluto 
2179 Pacific Avenue 
San Francisco, California 94115 
SHORT COMMENTS 
General 
Fiche is good for scanning and storing info where only occasional 
reference is needed. 
Paper copy needd for the few papers of 
frequently needed reference. 
It didn't even reach me until Jan. 5. probably due to a poorly 
registered address label. 
(I bet il lot don't make it if they're 
all as bad as mine.) 
I am a student and could not afford AJCL publications in printed 
form. 
You might make the microfiche envelope larger (to contain the 
opaque copy) and use face of envelope to print information on it 
You should have asked how long I was a member to interpret this 
stuff. 
Sorry this is late 
One of my major reasons for subscribing to 
AJCL is that it is published in microfiche. 
For $10 a year I am very satisfied. 
You might consider negative 
contrast to ease the eye strain 
I haven't found time to read articles but expect to and like the 
format. 
Love fiche. 
At this moment Itan not working directly in the field of CL. 
PLEASE NOTE: Negative, rather than positive, microfiche pro- 
vides the clearest and easiest to read copy. An option for 
negative fiche-copy would encourage a great deal of new member- 
ships. 
1) Do offer printed copies. 
than "prestige" (whose?) 
2) Currency is far more important 
Only major complaint about fiche is that all the readers are 
lousy on the eyes. 
I prefer to be forced to get used to microfiche because paper 
should not be wasted. 
Despite the current inconvenience for me of microfiche publica- 
tion, I am in favor of continued publication on the grounds that 
maybe it'll spread and eventually encourage better reader tech- 
nology. 
A lot of good work is being done by people who are not too 
famous; it's very importarlt that it should be published inexpen- 
sively somewhere. 
Question 1. 
Answer 1: But unrelated to microfiche as medium. 
Answer 0 : Because "go to inconvenienr place". 
Plan to buy a viewer as soon as I can afford one. 
Answer 6: Parts only. 
Question 2. 
Answer own a reader : Have 2. 
Department. 
Recently. 
Very recently. 
Costs about as much as a small B/W TV set. 
Answer go to a convenient place: Buying a reader. 
Answer none: But on the way to buying a reader or hopefully 
going to a convenient place. 
Question 3. 
Answer 6-10: All received 
Answer 11+: All. 
Question 4. 
Answer highly: Word "much" converted to "some". 
Answer hardly: On my job now. 
Question 5. 
Answer all checked: Papers not accepted submitted elsewhere. 
Question 6. 
Answer yes: But it gains others who need the pages. 
Most likely. 
But it is still experimental. Give it a chance. 
Loses some but gains others that are too long for 
publication elsewhere. 
Probably. Difficult ta be sure since it is a new 
j ournal . 
At least it would lose mine. 
But because of the format, gains others. Balance 
Answer no: 
Don't know but I suspect not since alternatives are 
so few. 
I feel there is more prestige in microform. 
Question 7. 
Answer yes: Inconvenient and costly. 
But not good. 
Probably, I don't really know. 
New one. I haven ' t used it yet , don ' t know how we1 $ 
it works. 
At cost. 
But only electrostatic process. 
Not clear or good 
type. 
At an inconvenient place. 
With inconvenience. 
Very difficult to access, though. 
Questisn 8. 
Answer 1: For use with students. 
Answer 2-3 : Good idea. 
This is the best offer yet--a postcard return to get 
paper copies of wanted articles only. 
No answer : Probably too expensive (postage) . 
Depends on the article content; question is not very 
good. 
Question 9. 
Answer fiche: With option of question 8. 
With option of #8, option of Finite String as part 
of this option, devote several cards of an issue 
to microfiche readers, their advantages and dis- 
advantages and cost. 
(To "mixed") NO! One format only. 
.Answer mixed: 
Fiche with option to order hard copy. 
Bit depends on which goes where. 
Answer other: 
EithPr fiche or paper but not mixed. 
Either fiche or mixed. 
No paper. 
Paper and mixed. 
Depends on mix. 
Question 10. 
Answer no: Would use a library copy and Xerox. 
Probably not. 
Offer on paper only. 
Answer ? Yes and no; pay more for the larger journal to 
continue; prefer microfiche. 
Stupid question--depends on question 8 situation. 
If paper copies of separate articles not available 
probably so--otherwise not. 
Question 11. 
Answer continue: Continue a "fairly" satisfactory method. 
Would buy "possibly if other material also available. I I 
Continue an "unsatisfactory" arrangement. 
Continue a-satisgaeeery arrangement for ... 
11 *' 
Continue an uX sntisfactory.. . 
Continue to "look for" satisfactory. . . 
Answer not read: I find microfiche inconvenient but do read 
articles of great interest. 
Answer other: Not read as much as I would if it, were on paper. 
11 
Question 12. 
The following are not suitable topics for ACL to publish on: 
Psychological applications social science applications, analysis 
of literature, instructional systems, programming languages, and 
information structures. 
TUTORIALS 
from the 
IEEE COMPUTER SOCIETY 
Text and References for Computer Net- 
works: A Tulorlal, June IH. 1975-?% pp 
Our biggest selllnr ..torial 1s now In ah 
~h~rd prlntlng. Top~c~ lncludc nauork con- 
flpurat~on and \ocabulary. hardwire compo- 
ncntb. softwarc components, ui~l~zat~an. 
meaiurcment and e~aluaaon, costs, and 
ISSU~S In management 
Prlcc-$12.00 
Mlcroprogramrning, Ma!. 17. 19?5-309 pp. 
Tev lncludes an introduc~ion by ~hc in- 
\rrucron. J. Michael Galey of IBM and 
Rlchard 1. Klelr of Browne & Ramarnoorthg. 
lnc.. and 19 papers on mlcroprogramrnlng 
In sit general categories: introductory papers. 
m~croprogrammed arch~tecture and lan- 
guages, production of mlcroprograrns. emu- 
lat~on and direct execution, microprogram 
applications other than instruc~ion execu- 
tlon, and conclusions and perspectives. 
Prlce-S12,00 Member price-S9.00 
Clolhbound-516.00 Clothbound 
member price-5 12.00 
Protection of Information in Computer 
Systems, Tutorial from Compcon 75 Fall, 
September 8.4 975-260 pp. 
The course covers the basic principles of 
information protection and explains the sepa- 
ration of addressing and protection, thc 
capati~lity sgwrm. [he acce~3-control list 
system, protect~ng barlous types of objec~s, 
and protected objects and sub\>\tsm\. 
Prlsc-S 12.00 Msmhcr price--$9.00 
Structured Programming, Tutorlal from 
Compon 15 Fall, Seprcmhcr H. -J'l75-. 
241 pp. 
Thl\ turor~al lncludcs an rnlroductlon lo. 
the major idea& of structured programrnlng. 
thcrr theoretical and practrcal foundalrons. as 
ucll a\ detallcd prcrcntallonb of a major 
appl~cation program \).%tern and a single 
program stre\slnp the top-dou11 evolutionary 
character of the procrsc. 
Price-S12.00 Member pr~ce-59.00 
Minicomputer Realtime Executiuer, Tu- 
torial from Compcon 74 Fall. September 9, 
1973-208 pp. 
Applicalions such as data acquisition. 
process control, communications, and bu4- 
nets data procaring are briefly discushed. 
Various reattime executives are discussed with 
emphasis on the way the choice of design of 
the ~arious key element\-memory rnanage- 
ment, scheduling, data rtructuring and file 
handling, interprogram communication, and 
control structures-affect the application of 
the sgrtem. 
Price-Sl5.00 
@ IEEE COMPUTER SOCIETY 
5855 h:ic Ci Plaza Sl;:tl; 30; Lon;: Bsach Callforrl~a 90803 Te'ephnn~ (213) 438-9951 
Name 
-- 
IEEE Computer Society Memb. No. 
.- . -. -.- 
Address - - 
.- 
Cit ~\State -- Zip, -. Counlry -----.- . 
Please send me the following tutorrats from the IEEE Gomputer Society 
Computer Networks copy (copies) at $ -- Structured Programming ..---copy (cop~esj at $ 
Mlcroprogramrning - copy (copies) at $ - _. Minicomputer Realtime Executives -.-copy ;copies) 
Protect~on of Information copy [cop~es) at $ .-- at $ -- --- 
Cal~fornia residents add 6% sales tax TOTAL %,-, .--- 
- ..--. - -.. - . .- -- - 
Be sure to state IEEE membership number if you claim membership discount. 
( I Mypaymentof is enclosed ( 1 Please send me your complete publicat~ons 
catalog. 
( 1 Add S2.00 billing charge & invoice me ( Please send lnforrnation on jolnlng'the Computer 
Society 
Send to: 
IEEE Computer Society, 5855 Naples Plaza, Suite 301, Long Beach, CA 90803 
MICROCOPY RESOLUTION TEST CHART 
NATIONAL 0UREAU OF STANDARDS 1964 A 

REFERENCES 
Fulton, D. L. A Plasma-Panel interactive graphic system. Proceedings of the Society for Information Displtay 15(2) : 74-80, 1~14. 

Stokoe, W. C., Jr. Sign Language structure: An outline of the visual communication system of the American deaf. Studies in Linguistics, Occasional Papers 8. University of Buffalo Press 1960. 

Stokoe, W. C. Jr. semiotics and human sign languages. Mouton, The Hague, 1972. 

D. dobrow, B. Fraser, "An augmented state transition network analysis procedure", Proc. 1st IJCAI, pp. 557-568, (1969). 

A. Colmerauer . "Les sys temes-q ou un formalis~e pour analyser et synthet iser d2~ phrases sur ordinateur", Project de Traduction automatique de llUniversite de Xontreal, TAD1 71, Jan., (1971) . 

C. J. Fillmore. "The case for case", in Bach and Harms (eds.), Universals in Linguistic Theory, Holt, Rinehart 5 krinston, pp. 1-90, (1968). 

C. Hewitt. "PMNKER: A language for manipulating models and proving theorems in a Robot", in Artificial Intelligence, Washington, D. C , May, (1969) 

?i. Xagao, J. Tsujii. '"~echanism of deduction in a question answering system with natural language input", Proc. 3rd IJCAI, pp. 285-290, (1973). 

Y- Sagao, J. Tsujii. Programing language for natural language processing - PLATOX", J.IPSJ, Val. 15, pp. 654-661, (1974). 

M. Nagao, J. Tsujii. Analysis of Japanese Ser tences by Using Semantic and Contextual Information". (forthcoming in AJCL-1976) 

V. Pratt. "A linguistic oriented programming language", Proc. 3rd IJCAI, pp. 372-381, (1973). 

J. Rulifson, et. al. QA4-A language far writing problem-solving programs", SRI Technical Xote 48, Kovember , (i970). 

J. Thcrpe, I?. Bratley, H. Dewar. "The syntactic analysis of English by machine". In Michie (ed.), Machine Intelligence 3, New York, American Elsevier, (1968). 

T. Vinograd. "Procedures as a representation for data in a computer program for understanding natural language", MIT Thesis, (1971) . 

W. Woods. Augmented transition network grammars for natural language analysis", CACM, Vol. 13, pp. 591-602, (1970). 
