~merican Journal of Computational Linguistics 
Microfiche g : 1 
THE FINITE STRING 
NEWSLETTER OF THE ASSOCIATION FOR COMPUTATIONAL LINGUISTICS 
VOLUME 11 - NUMBER 3 
DECEMRER 1974 
This issue was released for production an March 25, 
1975. The editor intends to distribute American 
Journal of Computational Linguistics in four packs 
per calendar year, promptly at intervals of three 
months. Each pack is to contain two numbers of 
the Finite String. The difficulties of the first 
year of publication of AJCL are responsible for 
the d'elayed, production of this ~ack , which also 
contains Volume 11, Number 4 of TFS. ~k would be 
a rash editor indeed who guaranteed promptness 
without caveat. The present editbr must warn the 
subscriber that'the end of the diLf.iculti-es is not 
yet fixed for a date certa.in. 
AMERICAN JQURNAL OF COMPL'TATIONAL LINGUISTICS is published 
by the Center for Applied Linguistics for the Association 
for Computational Linguistics. 
EDITOR: Da @ d G. Bays, Professor of Linguistics and of 
Computer Science, State University of New York, Buffalo. 
ZDITORIAL STAFF: Brian Phillips, Assistant; Jacquin Brendle, 
Secretary. 
EDITORIAL ADDRESS: Twin Willows, Wanakah, New York 14075. 
MANAGING EDITOR: A. Hood Roberts, Deputy Director, Center 
for Applied Linguistics. 
ASSISTANT: Nancy Jokovl ch . 
PRODUCTION AND SUBSCRIPTION ADDRESS: 1611 North Kent Street, 
Arlington, Virginia 22209. 
Copyright 1975 by the Associatiom for Computational L;inguistics 
THE FINITE STRING 11-3 
INTERNATIONAL CONFERENCE - COLING 76 . . . . . . , . . . . . 3 
WORKSHOP - theoretics-1 issues in natural language processing 4 
SUMMER SCIiOOL - computational linguistics at Rocquencourt . . 10 
SUMMER SCHOOL - literary statistics at Cambridge, England . 13 
INTERNATIONAL CONFERENCE - computers and the humanities . . 14 
nfai.s CONFERENCE - information interfaces . . . . . . , . . 19 
IEEE CONFERENCE - computers to reach the people . . . . . . . 21 
IFIP CONFERENCE - computers in education . . . . . . . . . . 22 
INTERNATIONAL CONFERENCE - developing countries . . . . . . . 23 
NSF DEPUTY DIRECTOR .L. Richard C, Atkinson nominated . , . . . 24 
NEH CALENDAR - grant applications . . . . . . . . . . . . . . 25 
NATIONAL PROGRAM DRAFT - library and information services . . 26 
RESEARCH PROGRESS - Index Thornisticus in press. . . . . . . . 26 
COMPUTER SECURITY - AFIPS and ACM publish guides . . . . . . 27 
OPINION: A restricted sublanguage approach to high quality 
translation - V~ctor Raskin . . . . . . . . . . . . . . 28 
TECHNIQUE: Letters with variable values and the mechanical 
inflection of Rumanian words - Nlnerva ~ocsa . . . . . . 38 
ACL: Secretary-Treasurer's report . . . . . . . . . . . . . 53 
Editor's report . . . . . . . . . . . . . . . . . . . . 54 
American Journal of Computational Linguistics 
Microfiche 9 : 3 
COLING 76 
INTERNATIONAL CONFERENCE ON COMPUTATIONAL LINGUISTICS 
The conference originally planned for September 1975 has been 
rescheduled in order to avoid conflict with meetings on 
artificial intelligence (the date of that meeting was changed 
after the CL date was first set) and on applied linguistics. 
The general plan is as described in ACJL; Card-6. 
The coordinator is Dr. Guy Rondeau. 
A request for further information should contain Title, name, 
and surname; Post held; Department; InStitution: Postal address 
and zip code; and Field of interest. 
Addresq COLING 76, Linguistics, University of Ottawa, KIN 6N5. 
American Journal of Computational Linguistics 
~icrofiche 9 : 4 
INTERDISCIPLINARY WORKSHOP 
THEORETICAL ISSUES IN 
AATURAL LANGUAGE PROCESSINGo 
MASSACHUSETTS INSTITUTE OF TECHNOLOGY 
Sponsored by the Association for C~mputational Linguistics 
Supported by the Mathematical Social Science Board with 
funds granted by the Nathal Science Foundati~n 
DIRECTORS : BONNIE NASH-WEBBER AND ROGER SCHANK 
The announcement and program appear on Card 25. The fol- 
lowing account is presented as a sratement bf the needs and 
condition of the field; it was prepared by BNW and RS. 
The workshop is to promote interaction among researchers in 
computational linguistics, psychology, linguistics, and artificial 
intelligence. The primary purpose of the workshop is to explore 
two major areas of common research, memory and knowledge, both 
with respect to their relatiofi to human language behavior. Spe- 
cifically, we hope to consider such questions as: 
(1.) What computational models and mechanisms have been 
proposed up to now in these areas? 
(2) What aspects of huma~ language behavior are they 
meant to account for? 
(3) Are these models and mechanisms compatible? 
(4) Is there a single global view of language understanding 
and use that is adequately modelled by some combination 
of them? 
THEQRETICAL ISSUES 
(5) Are there still significant aspects of human lan- 
guage use which they cannot account for? 
(6) What is the best model of human lanquage use that 
can be assembled out of the concepts that nave been 
developed in computational linguistics, linguistics, 
P 
psychology, and artificial intelligence? 
(7) How well does it really -approxima%e what humans do 
with language? 
(8) With respect to gaps in the model, is thefe any- 
thing currently in the wind-adequate to complete 
them? 
Thus our primary purpose is both to provide an assessment of our 
mowledge and goals in this area .with respect to language pro- 
cessing and to provide a cross-disciplinary tutorial for the par- 
ticipants. A secondary purpose of the ~orkshop is to discuss the 
important issue of valid methodology in such research. 
In eight ses:sisions, speakers will address the audience on 
points made by themselves or others in posimtioh papers pxeviausly 
distributed to all participants. After these presentations are 
finished, speakers and audience will take part in general dis- 
cussions. 
MOT I VAT1 ON AND GOALS 
In recent years, researchers in several disciplines have been 
converging on the problem of language understanding and memory as 
providing a handle on the problems in their own fields. Research- 
ers in Artificial Intelligence, conce.rnad with building models of 
intelligent behavior, have.started to develop and study models of 
conversational interaction, which naturally rely on models of 
language understanding. This is often done without reference to 
the work of researchers in other disciplines. 
Lingui,sti-cs, of course, has always been concerned with lan- 
guage, but "frequently at no level higher than that of the sen-. 
fence. Such theories as derived fkom the consideration of sinale 
sentences out of context are diffioult to apply to the task of 
understanding. Recently*some linguists such as Chafe have 
shifted thelr emphasis towards understanding and modelling dis- 
course, which is af direct relevance to questiorls of understand- 
ing. 
Psychologists have also been interested in problems of lan- 
guage understanding and memory. While early research did look to 
l~.nguistics to p-ovide plausible theories of human language be- 
havior 
the then-curtent theories proved insufficient to the task. 
As a result, some psychologists have begun to create their own 
theories, while others have begun to pay attektion to the wark of 
A1 researchers. Computer models are now appearing which are ex-. 
pli-,cations of these theories: 
In the field of computational linguistics, the challenge of 
building computer systems which can carry on fluent and helpful 
dialogues with a user has dlso shifted the emphasis in the field 
from more efficient parsers to mote capable understanders, 
In spite of this convergence of many different disciplines 
on the same problem, there is sqrprisingly little communicat.ion 
between researchers in the different fields, apart, from occasion- 
al discovery of nrelevant paper$ in one field by members of anm- 
other. Without conferenqes of the sort that we are proposing, 
there is. no opportur)ity for the individual researchers in these 
different fields to talk to each othek directly. Th'e value of 
SUCR face-to-face confrontation and the opportunity for asking 
questions and sxploring tne applicability of techniques in areas 
other than the ones that the author had in mind are well appre- 
ciated. Wjthin disciplines such opportunities exist in tradi- 
tional professional conferences. Our conference will provide an 
opportunity for such interaction across a diversity of fields 
which would not otherwise be possible. 
THEORETICAL ISSUES 
The need for such a conference' is especially great in the 
area of natural language understandlnu. 
The many different fields 
whlch are beginning to glve Strong attention to this problem all 
havz different interests and consequently different emphases on 
the problem. 
Researchers in a particular field tend to focus 
only on their own interests and ignore other aspects of the prob- 
lem. For example, the classical transformational grammar theory 
has largely ignored the necessity for the theory to account for 
psycholingulstlc and other pefformance aspe~ts of language. 
This 
is a reasonable way t~ gain a restricted research problem, but the 
result of such research mag suffer if the researcher does not have 
some general idea of the problems associated wsth the aspects that 
are being ignored. Thls conference will attempt to provide specl- 
allsts in different fields with this type of general understanding 
of the problems of concern in other flelds. We feel that thrs ex- 
posure to different aspects and enfphases will have a very benefl- 
clal effect on all flelds of natural language research, and that 
wlthout such'interchange the potential for much of that research 
will not be reallzed. 
The opportunity for duch a rneetlng as we are preparing does 
not currently exlst elsewhere. ~onferencbs sponsored by profes- 
sional socletles invarlabky present intradlsclpllnary news rather 
than interdlsclpl~nary. ones, and past interdlsclpllnary workshops 
have always been on a very small scale. For example, at the NSF 
spoqsored workshop in Woods Hole, Massachusetts, in 1971, some 25 
researchers interested in Computational Semantics were brought 
together for two weeks of worthwhile talks. Wh~le the results of 
such a conference are largely intangible, a number of partioipants 
In that oonference have attested to the lmpact of that conference 
on thelr way of thlnklng about problems aqd the course~of thelz 
research. 
We plan to adhere to the model of the Woods Hole cob 
ference, but with the following two changes. 
THEORETICAL ISSUES 
First 
it is important to creatb the possibility fol: the 
many new researchers from different 'fields who have entered this 
area to gain an appreciation of the different emphases of other 
fields. Secondly, it is important that a much larger number of 
people whose interests are .in one discioline or another be ex- 
posed to the Ideas emerging from the synthesis of these di-sci- 
plines. It is important that more than just a small group 'be 
able to exchange ideas. 
ence 
ing 
Since a long workshop would be very 
re proposing, we are relying on an e 
papers to familiarize all participa 
alike, with the current ideas. on na 
in each of the fields. These prepri 
diffic 
arly ci 
.nts, sp 
tural 1 
nts wil 
ult on 
rculati 
eakers 
anguage 
1 also 
the scale 
on of ~osi 
and audi- 
understan 
serve the 
valuable job of informing those unable to participate in the work- 
shop of these current iaeas, and de therefore intend to make copies 
of the preprints widely available. 
ORGANIZATION 
The two sessions of each day will be held in the morning.and 
in the late afternoon. The long break for\ lunch in between wfll 
facilitate discussions 04 the morning's topic, without the need to 
get back to another session immediately. 
All sess'ions will be open to the public, and we expect about 
I50 people to participate. MIT was selected as a site so that 
the widest group of interested people. might be able to come at a 
reasonable cost. MIT has made a large air-conditioned lecture 
room available, and will also provide low-cost dormitory housing 
for the participants. 
Ihe sessions will-not simply be introductions to working 
systems or wel-1-known theosaies . Posit i on papers will have been 
distributed to all' participants at least a nonth in advance of 
THEORETICAL I-SSUES 
the workshop to f~amiliarize them with the ideas of aach speaker. 
The sessions will consist of short presentations (1'0-15 minutes) 
by the speakers outlining their already circulated ideas. Fol- 
lowing a break, each speaker will have the opportunity to respond 
to earlier remarks, after which discussion will be opened to the 
audience. A session chairman will bB responsible for maintaining 
the Level and direction of the sessim. 
American Journal of Computational Linguistics 
Microfiche 9 : 10 
SUMMER SCHOOL 
COMPUTATIONAL LIliGUISTICS 
DOMAI NE DE VOLUCEAU - ROCQUENCOURT 
Sponsored by the Institut de Recherche dlInforma tique et 
d 'Automa-tique 
1 NTRODUCTI ON 
Generalities about the methods, problems, and applications of 
computational linguistics. (Level of analysis of content, docu- 
mentation, indexing, aids to diagnosis ,programmed instruction in 
natural language, etc. ) Is automatic analysis of language pos- 
sible? Limits and possibilities of results. Possible applica- 
tions, justi£.ication. 
MATHERATICAL METHODS IN LINGUISTICS 
The essential components of the structure of language (French as 
the example. Usable strategies in the automatic analysis and 
production of text. Difficulties encountered. From language to 
algorithms to programming. Does language have a mathematical 
structure? How does it come out in French? ~ifficulties in the 
choice of methods of analysis and production. Adequacy relations 
between natural and programming languages. 
Presentation of an operational discovery procedure which, begin- 
ning with a corpus analyzed grammatically, makes it possible to 
obtain automatically a syntax allowing disambiguation. It is 
impossible to foresee all the peculiarities of language. One 
must therefore arrange to integrate new linguistic data as they 
arrive; for that a discovery procedure is necessary. 
ROCQUENCOURT SUMMER SCHOOL 
AUTOMATIC DOCUMENTATION 
A mathematical model of content analysis, used in automatic in- 
dexing and in interrogating documentation systems. 
Presentation 
of realizations. 
Automatic documentation is a privileged field 
of applicatkn for the most advanced methods. 
It is equally a 
particular viewpoint from which to see linguistics. 
The first part of the course will be treated jointly by 
M. Andreewsky and M. Fluhr, 
USE AND PROOF OF THEOREMS IN LINGUISTIC AND INFORMATIC APPLICATIONS 
!I, Pitrat, C.N.R,S, 
AUTOMAT1 C CONTENT ANALYSIS OF SCI ENTI FI C TEXT WRITTEN IN NATURAL 
LANGUAGE 
M. Daniel Herault, Universite Pierre et Marie Curie 
Discursive double articulation of scientific discourse: the 
hypersyntactic and hypersemantic components. Definition of seman- 
tic content. Role of the underlying derivational system; at the 
Level of a text, in the realization of the double articulation: 
semantic units (predicates), principal modifiers, and associated 
syntactic structures. Informatic realization for the Slavic lan- 
guages. Remarks on German, Romance, and Japanese. Elaboration 
of an advanced documentation system: integration of this research 
in an MT system. 
AUTOMATIC TRANSLATION 
M. Yauqtlois, GETA ~renob2e 
Automatic translation by syntactic analysis: 1960-1970. The 
process of translation as the step following source analysis, 
followed by transfer to the level of surface syntax. Notions of 
structural descriptors and their diverse representations in lin- 
guistic schools. Types of grammars and algorithms which permit 
automatic analysis or production. Evolution of different strate- 
gies aimed at new approaches to.MT. Levels of transfer in MT: 
Surface syntax (Japanese experiments); transformational grammar 
(American experiments) ; pivot languages (Grenoble experiments; 
Mel'chuk-Zholkhawski theory); critique. Current research in MT: 
Practical work (machine aided translation--pre-editing , revising 
short and middle-term possibilities); long-term research- (aspects 
of semantic calcu-lus; experiments in man-machine communication 
in question-answering systems; influence of semantic research on 
automatic translation). 
ROCQUENCOURT SUMMER SCHOOL 
MAN-MACHINE DIALOGUES AND SPEECH 
M. Gueguen, E.N.S.T. 
Objective analysis of language: physical structure of the sig- 
nal of language, classic methods of analysis, analysis by rnodel- 
ing (linear prediction, analysis, by synthesis) . Automatic 
recognition: system organization, acoustic preprocessing, 
levels of recognition, use of linguistic data; realizatinns. and 
Qpen problems in automatic comprehension of speech. Speech 
synthesis: devices and their commands (vocoders with channels, 
formants, simulation of the vocal tract); levqls of synthesis;' 
synthesis by rules. Perspectives and conclusions.: the help of 
computational linguistics in the area. 
ROUND TABLE : COMPUTATIONAL LINGUI ST1 CS AND LI NGUI ST1 CS 
American Journal of Computational Linguistics 
SUMMER SCHOOL 
LITERARY STATISTICS 
CAMBRIDGE UNIVERSITY, ENGLAND 
sponsored by the Association for Literary apd Linguistic Computing 
-- - , U..-U 
. 
ACADEMIC SECRETARY 
PRINCIPAL LECTURER 
GUEST LECTURERS 
LECTURE TOPICS 
TUTORIAL TOPICS 
ONLINE FACILIT'IES 
INFORMATION 
F E-ES 
M. H. T. Alford, Esq. 
2, Sidgwick Avenue 
Cambridge, 13nglan$ 
Norman Thomson 
IBM and Southampton University 
H. Sykes-Davies , Cambridge 
A. Q. Morton, Edinburgh 
Ye T. Radday, Haifa 
R. W. Balley, Michigan 
K. W. Kemp, Cardiff 
Estimation and confidence intervals 
Design of experiments 
Analysis of var,iance 
Exercises based on the lectures 
Statistical validity of the work carried out 
Practical demonstrations 
Available students 
a1nose who inform Mr. Alford of their h~pe 
o£ attending will recei ve 'further infor-ma- 
tion. A tentative reservation of living 
space is suggested. 
Tuition L20 for nonmembers, L17 membela 
Accommodation about L30; room and all 
meals, from dinner 7/13 through breaEfast 
7/19 - 
American Journal of Computational Linguistics Microfiche 9 : 14 
CORPUTERS Ailb THE tiURA!iITIES 
UN I VERS ITY OF SOUTHERN CALI FORN IA 
ORGANIZING COMMITTEE 
Robert Dilligan 
407 Founders Hall, USC, Los Angeles 98Q07 
Rudolf Hirschmann 
- 
Joseph Raben Queens College, CUNY 
Donald Ross University of Minnesota 
Todd K. Bender University of Wisconsin 
Grace C. ~ertlein California State ~nfversity , Chico 
ADVISORY COMHITTEE 
John R, Allen, Riclard Bailey, Emmett Bedford, Roy Bo~~s, 
T. H. Howard-Hill, Winfried Lenders,  illy Martin, Joan Smith, 
John B. Smith, F. de Tollenaere, Stephen V. F. Waite, 
Stanley N. Werbow, Roy A. Wisbey, Antonio Zampolli 
EDITORIAL NOTE 
A tentative list of contributions appears on the following 
frames. It is to6 late to stimulate attendance, but it shows 
the scope o'f current activities and may sGggest .further 
exchanse of information. 
TENTATIVE PROGRAI", 
Raymond Erickson 
Queens College, CUNY 
Bo Alphonce 
Yale University 
Gary Nelson 
Oberlin C~nservatory 
Fred T. Hofstetter 
tniversity of Delaware 
Lynn Trowbridge 
University of Illinois 
Norbert Baker-Meil 
Staat. Inst. Musikforsch. 
Fred T. Hofstetter 
University of Delaware 
Michael J. Ramey 
UC LA 
VISUAL ARTS 
Luraine Tansey 
San Jo-se Corn. College 
Eleanor G-uralnlck 
The; DARMS project 
INTRIX: A scdnner for pitch-class 
pat terns ia multipart music 
Jk formalization of musical syntax 
National differences and similari ties 
in, the use of melodlc intervals ,during 
the mid-19th to early 20th centuries 
On the analysis of musical analysis 
A computer processing system for 
Renaissance ousi c 
.A computerized lndexing of Renaissance 
rnusi c 
De~~elopment of a center for computa- 
tional musicology 
Computer application to the cornpara- 
tive study of musical instruments 
Pre-~ol umbian art: si ces and chrono- 
logy compu erized 
The proportions of Archaic Greek 
sculptured figures: a computer study 
Iiiroshi Xawanc Markov process theory of plctures 
Elizabeth M. Lewis 
West Point 
Computer cod,ing for a rnlcrographlc 
index in art 
James E Gips & George N. Stiny Computer models for 
CCLA aesthetics 
Leonard 3:evers a Computer animated film as visionary art 
Calif St. 
COKPUTER-ASSISTED INSTRUCTION 
Tej Bhatia New directions and issues in CAI 
University of Illinols 
Robert L. Oakman A videotape course for computer 
University of S, Carolina education in the humanities 
Peter ZoUer. d CAI 'approach to Black Engl-ish 
Wichita State University 
Francine Ouellette JEUD~MO: A practical workshop 
University of ~ont~rkal 
George 0 ' Brien A: Siren songs and a skeptlk 
University of Minnesota 
DICTIONARIES AND CONCORDANCES 
Sarah K. Burton Hunter Evolution of lanqbages, Part I; 
University of Xlaba~a Romance etymplogy 
Sidney Berger 
UC Davis 
Compiling a concbrdahce 
Johannes B. Casser  he, Index Thomisf icus: A test-case 
University of Montrreal 
Robert Benson 
UCLA 
A prbposed computer concordance of 
Medieval Latin 
Andrew 2. Crosland  he concordance and the study of the 
U. S. Carolina, Spartanburq novei! 
Donald MR. Lance The use of the computer In dete,rminlng 
~niver$ity of Missouri the geographical distr~bution of Items 
Edward A. Klineb Computer applications in Middle 
University of Notre Dame English dialectology 
Michael M. T. Henderson Use of an in'teractrvc program ~n aria- 
U.  isc cons in, Madi'son lyzi~ng data fo'r a dlolett dictionary 
Paul, Bratley 61 Serge Lusignan Some problems and solut~ons in 
University of Kontreal the edition of a dictionary 
Richard W. Bailey Inter-actiye 1ex~ coyfaphy : Some uses 
university of Michigan of Michiaan Early Fodern Enclis-h 
~at~erials 
DATA BASES 
Vincent J. Ryan Cornputeri zed concurrent indexing 
UCLA 
Charles Dollax Scholars, computers, and the Fationdl 
Nat. Archives Records ~rchives 
LINGUISTICS 
Gerard Salton 
Cornell University 
Annette Paguot-Maniet 
university of Lava1 
Dirk Geens 
AvTL 
Patricia Lang 
SWRL 
Jean-Guy Meunier 
U. Quebec, Montreal 
Edward R. Ga~m~n 
Calif. St. U., Fresno 
Burghard B. Rieger 
Tech. Univ. , Aachen 
Robert A, Ariew 
Pennsylvania State U. 
On the role of words and phrases in 
the automatic content analysis of, texts 
Le vocabulaire caracteristique de 
l'avare chez Plaute et chez koliere 
Automatic syntagrnatic anal ysls or 
English 
L.A.P.: A system for processing text 
A system for interactive text pro- 
cessing and content analysis 
Nugerical taxonomy in linguistics 
On a toi-erance-topology model of 
na tura2 language semantics 
Andre Breton's Poisson soluble: A 
computer-aided study 
Jay Leavitt & John' Lawrence Mitchell Gap recurrence: d lexico- 
University of ~iinnesota statistical measure 
David Sankoff Correlates of speakers' word frequency 
University of Montreal 
I 
parameters in a corpus of sgoken French 
Barr~n Brainerd On the distributions of articles and 
University of Toronto pronouns 
TEXTUAL ANALYS I S 
Robert Cannon An optipnal text collation algorithm 
University of S. Carolina 
Todd K. Bender Ar literary work conceived in positional 
U. of Wisconsin, Madison notation 
Giorgio Buccellati 
UCLA 
Computer aided analysis of Cunei farm 
texts 
Eric Poole 
The coziput'er in textual  collation and 
Ueiversaty of Kent 
stemmatic analysis 
STYLISTICS 
Richard Williams Diction and social class iq 
Ivichita State .University seventeenth century Spanish drama 
Colin E. Martindale 
University of Maine 
The Night Jourqey: Pdtterns df re'- 
gress-ive imagery in journeys to Hell 
Tommy Joe Ray Thehe as stqbe 
University Q£ ~lississippi 
James-Joyce 
Uc Berkeley 
Donald Ross 
University of Minnesota 
John Odrnark 
~niversitdt Regensburg 
David H, Chisholm 
University of Arizona 
Computational model ,of \stanzlaic 
patterns in English 
Keats' odes and sonnet--style and 
genre 
Computers and styJ*ist1c analysis 
PhonologicaJ patterning in German 
verse 
Geoffrey J. D. E. Archbold Repetition, a characteristic of 
university of Victoria Ammianus Marcellinus.' style 
Stephen Waite 
 artm mouth College 
Pierre Laurette 
Carleton University 
Daniel L. Greenblatt 
University of Missouri 
Effects of genre and some stylometri c 
features: evide'nce from Cicero's works 
La petite liseuse do poem 
automate @e lecture/r6&cri ture 
Variable rules and Literary style 
American Journal of ComputationaI Linpis tics 
~icrofiche 9 : 19 
GONFERENCE COMMITTEE 
Befir. H. Weil Exxon Research and Engineering Co. 
Joseph Coyne National Technical Information Service 
Anh Farren BioSciences Information Service1 
A. Hooa Roberts Center for Applied Linguistics 
BIBLIOGRAPHIC CONTROL 
Ellis Mount Bibliographic standards work-. 
Columbia University. natidnally 
Eric Clyde Bibliographic standard's work-- 
Canada Inst. S-T Information internationally 
Lawrence bivihgston CONSER ~roject 
Council on Library Resources 
Maureen LeFever 
BIOSIS 
BIOSIS/CAS/E~ bibliographic guide 
for authors and editors 
USER ASPECTS 
Juey Wanger 
Impact of on8-line bibliographic 
Syster Development Corp: services--a preliminary report 
Colin K. Mick 
Impact of on-line search services 
Applied Communication Res. 
on public library operations 
James L. Carman 
Roles of intermediary and users 
University of Georgia 
in bibliographic retrieval sys.tems 
REPAC~AG I NG OF ABSTRACTS 
Irving Zarember 
her. Petrolew Institute 
James Gape 
Energy R&D Administ~ation 
I 
AYl /DLKWEiVT "Patent Alerts" 
ERDA Bibliographic Data Base 
NFAIS 
DOU~~~S A. Fisher 
BIOSIS 
HGEP 
MI LES CONRAD MEMORIAL LECTURE 
Melvin 6. Day Sharing--the hope of the 
National Library of Medicine seventies 
DOCUMENT ACCESS 
James L. Wood 
nfais member services study 
Chemical Abstracts Service report 
Margaret H. Graham API-CAIS experimental metropoli- 
Exxon Res. & Eng'g CO. ta-n library service 
Paul Zurk~wski Business implications 
Information Industry Assoc. 
Roger Summit On-line ordering of docum4nts 
Lockheed Retrieval Service 
American Journal of Computati~na ~inguistics Micr6fiche 9 : 21 
COMPUTER TECMNOLDGY TO REACH THE PEOPLE 
PROGRAM EXTRACTS 
Martin L. Rubin 
HumRRD 
Susan Wittig 
University of Texas 
Austin 
Kerry Mark Jo81s1 
Anes Research Center 
Lister Hill: A national CAI 
network 
CAI in the composition c~a.ssroom: 
some practical ansuers and some 
philosophical problems 
The megauniversity of Athens: 
A scenario for the future 
0. Firschein 61 R. K. Summat Computerized retrieval in 'tl 
Lockheed Information public library setting 
Systems 
American Journal of Computational Linguistics Microfiche 9 : 22 
COIIPU-TEFS IN EDUCATIOH 
TO BE HELD UNDER THE HIGH PRESIDENCE OF THE 
Under the patronage of UNESCO, OECD, and the Commission of 
European Communities; with the assistance of the Direction 
~Qne'rale de llIndustrie and the Intergovernmental Bureau *for 
Informatics; in cooaeration with the International Commilssion 
on Mathematical ~nsiruction and the International Commission 
on Physics  ducati ion. 
Organized by the Association Franyaise pour la Cyberndtique 
Economiquq et Technique 
REG I STRATI ON 
Registration is 500 F. until May 1; 600 Fa thereafter. Write to 
AFCET - B.P. 571 - 75826 Paris CEDEX 17 for forms and details. 
-Wenty papers have been ingitied; 740 contributions have been 
submitted to referees. 
Information xbout the scope of the conference was published on 
AJCL Card 8. 
American Journal of Computational ~iaguis~ics 
Microfiche 9 : 23 
CONFERENCE ON NATIONAL PLANNING FOR INFORKATICS IN 
DEVE LOPIIIG COUilTPIES 
For full information: 
Dr. H, A. Al-Bayati blr. A. A. M. Veenhuis 
~irector' General Intergovernmental Bureau 
National Computers Centre for Informatics 
P.0. Box 3261 - Saadoon E.O. BOX 10253 
~aghdad, Iraq 00J44 Rome, Italy 
American Journal of Computational Linguistics 
Microfiche 9 : 24 
NEW DEPUTY DIRECTOR FOR 
NATI ONAL SCI ENCE FOUNDATION 
Richard C. Atkin,son, Stanford psyczhologist, is President Ford's 
nominee, according to a March 4, 1975., release. 
Atkinson, a creative designer and user of mathematical models 
for memgry, learning, and behavior, is assistant dean of the 
scho~l of Humanities and Sciences at stanford andvhairman of 
its Psychology department 
A member of the National Academy of Sciences, National Academy 
of Education, and the American Academy of Arts and Sciences, 
Dr. Atkinson has written or edited ten books and more than a 
hundred professional papers. He was educated at Chicago and 
Indiana, and has taught at UCLA and Michigan. 
American Journal of Computational Linguistics 
Microfiche 9 : 25 
NATIONAL ENDOWMENT F~R TH~ HUMAN I TI~S 
CALENDAR OF APP-LI CAT1 ON DEADLINES 
SPRING AND EARLY SUMMER 
May 9 Research Grants Beginning a'ftfter January 1, 1976 
Sirnone Reagor, Division Director - 202-382-1072 
Six programs : Research tools (dictionaries, biblio- 
I 
graphies, guides, and catalogs) . Research centers 
(major research collections with topical focus) . 
Internation~l conferences for the Bicentennial (the 
deadline for this program is past). State and local 
history (scholarship and archives). Editing (of his- 
torical and literary papers of scholarly value). 
Mav 12 Fellowshi~s 
For 1976-1977 
James Blessing, Division Director - 202-382-1491 
Independent study and research for scholars, teachers, 
writers, and other interpreters of the humanities who 
have produced or demonstrated promise of producing 
significant contributions to knowledge. Six months 
(to $10 ,d00) or twelve months (to $20,000). 
June 26 Public Programs Beginning after December 15 ,' 1975 
John Barcroft, Division ~irector - 202-382-1111 
Museum personnel program. University or -internship 
programs, seminars, or workshops to train interpreters. 
July 1   ducat ion Programs Beginning after Jan~ary 1, 1976 
Roger Rose~blatt, Divi-sion Director - 202-382-5891 
Procjrhm grants for critical re-examination of the 
content, orgahization, and method of presentation of 
a group of related courses or an ordered program of 
study in the humanities. The central topic can be a 
region, culture, era, etc.; or a program can be 
defined by a curr-icular level. Limit, $180,000 in 
three years. 
American Journal of Computational Liqguistics 
~icrofiche 9 : 26 
REV1 SED DRAFT 
A ~~ATIONAL PROGRAH FOR LIBRARY AND INFORMATI ON Y ERVI CES 
National Comnission on Libraries and Information Science 
Suite 601, 1717 K Street NW Washington 20036 
The conunlssion expects to introduce draft Federal legislation 
during 1976. The draft reflects comments received in letters 
and obtained through regional hearings and the professional 
press, 
INDEX THORISTICUS 
Roberto Busa, S:J. 
Fondamente NoVe 4885 
30121 Venezia 
Italy 
Progress during 1974: 
32 volumes of the Index Thomisticus, 
23 volumes of the Concordantia Prima, and 9 volumes of the 
Indices Distributionis, making 36,000 pages in all, photucom- 
posed. "Photocomposit'ion time was 60 second per page: slow 
but perfect!'' according to Father Busa. Ten volumes have been 
printed, bound, and published. Some 20 to 25 volumes are 
still to be prepared. 
American Journal of Computational Linguistics 
Microfiche 5 
SECURITY 
AFIPS SYSTEM REVIEW MANUAL 
The first of a series on system improvement 
Edited by Robert L. Patrick; based in part on a review of 
likerature conducted by Nary Elizabeth Stevens. 
The object is to specify methods of gathering data so that 
computing center managers, auditors, 
and system designers can 
assess their security needs. SlO.00 from AFIPS Press, 210 Summit 
Avenue, Montvale, New Jersey 07645. 
ACM EXECUTIVE GUIDE 
A booklet for executives and managers--those to whom EDP 
managers report, has been prepared by the Institute for Computer 
Sciences and Technology of the National Bureau of Standards and 
the Association for Cotnputing Machinery, with financial assistance 
from the National Science Foundation. 
Why? A man substituted deposit slips, magnetically coded with h4s 
account number, for the Hank ones available on a bank's 
customer counter.,... 
Who? programmer, janitor; or even manager.... 
Can data in a computer system be completely8protected? No. 
Termhals are the least secure points. 
The booklet lists technical and managerial solutions to 
partially protect against thege and other problems. 
Dennis .K. Branstad and Susan K. Reed 
Svstems and Software Division 
Insti tute for Computer Sciences and Technology 
Nation41 Bureau of Standards 
Washington, D. C: 20234 
American Journal of Computational Linguistics 
~icrof iche 9 : 2 8 
A RESTRI CTED SUBLANGUAGE APPROACH TO Y IGH QUALITY TRANSLATION 
Victor Raskin 
Institute of Philosophy 
Hebrew Uni versi t'y 
Israel 
This paper deals with an approach. to the problems of automa- 
tic high quality translation and, more generally, of automafic 
language data processing, based on the restriction of the input 
of MT and other systems to 3 certain type of suhlanguac/e. 
The 
approach was proposed by then present author in the framework of a 
general theory of sublanguages (see Raskin, 1971) and subsequent- 
ly tested and used by his own and other groups of researchers in 
the USSR on tne, material of diverse restricted sublanguages (see 
ibid; Gorodeckij and Raskin, 19711 R.1). The paper consists of 
two parts. Part 1,cpntains a very brief expositionc the basic 
principles of the approach. In Part z some advantages of the 
approach over other (unspecified) approaches unrestricted in this 
way are mentioned in the context of a few important problems of 
high quality translation. Since these problems were also dis- 
cussed by the contributors to Feasibility Study on Fully Automa- 
tic High Quality Translation, &t certain points of Part 2 the 
paper enters a dialogue ofd a sort with some of then1 (all the quo- 
tations and references followed by a name only are taken from-the 
coiitribution by the corresponding author to the said report). 
RESTRICTED SUBLANGUAGES 
1, 
SOME BASIC PPINCI PLES OF THE PESTRICTED SUBLANGUAGE APPROACH 
It might'be observed that in rrost cases when the practica; 
neecl of constructing an MT system arises, its input, 1.e. the 
lincuistic material which is to be subjected to such treatment, 
is highly restricted 4y certain conditions: 
it is usually a rela- 
tively narrow field of science or kechnology with texts which are 
relatively, hcmogeneous, with a limited vocabulary, a restricted 
set of Byntactic constructions, a highly structured substance of 
the content plane, and a relatively constant system of values for 
ail the re-letrant pragmatic parameters which aredetermined in 
this case not by the individual prwerties of any -partdculal; sit- 
uation of comunication, as is usually the case in casual com-.unl- 
cation, but rather by the position of che field itself among the 
contiguous fields as veil as in nor?.-linguistic reality, in general. 
For sucB restricteci sutlanguaces a simple algorithn of autom'atic 
processing was constructed and proved to be hig!~ly efficient 'in 
its practical applications. 
The algorithfi is based on an over-i~portant property whicil 
follo\:s, logically and practically, from the features which charac- 
terize the class of restricted sublanguaaes in the theory of sub- 
languages, including tilose which were emphasizes above and \~hi'ch 
result in the irrelevdnce of all surface srructure distinctions 
among sentences with identical deep structure or the exact syno- 
nymity of all-paraphrases (and, in fact, even near-paraphrases!. 
  his proljerty implies that each stem in the- vocabulary bf e re- 
stricted su61awuace tends to pla;,- 
certain permanent role in 
RESTRICTED SUBLANGUAGES 
a11 the situations described by those Sentences where t!ic sten 
occurs, no nlatter t~hetl~er it tal;es the ior~ of a vcr11, nwn, or 
any other part of speech. A riinimal sufficient inventory of 
these roles, wl~ich arc qiven the status 01 sernntic characteris- 
tics of stems, is conpiled (usual 1y It does not eAcecd 15' iters) 
and each dictidnary entry is assiqncd a ccrtain characteristic. 
Then a schere of the maxi~ally extended scn'tcncc of the restric- 
ted suhlancuage, a maximal deep structure of a sort, is postulated 
in, such a k:ay that each sentence (or rather, eat?] clause) can be 
represented as a (partial) realization of this structure. Such 
structures can ernl~ec?, nest in, erc., each other. The dictionary 
of the restricted sublanguagc wjth all its entries beinc assigned 
serantic ~Jlaracteristics ?nd the sc?er.e of the rr;l~ir.:all~~ extended 
sentence of the restricted sublanyuaue are t1.e two instrunents cn 
which, the univcrs.al-alqoril is founded. Texts of the restricted 
sublanquaqe constitute its inpuk, the cutput Leinq a sequence of 
(2artially) filled, ordered. and su1)ordinated sc!iecles/deep struc- 
Cures. By nakinq the suvantic characteristic asnapncd to eacl 
sbe~ of the restricted sublancuape, pore or less CctailcC, one ray 
cdntrol the depth of semantic analysis. ILt5 its sukalqorithr:~ of 
f 
''e1.lipsi.s analysis", "l~o~~oc~cneous parts analysis , "li~undar-~ anal- 
ysis", the algorith~ y~erates as a uhivcrsal Turin? r-,!c?:inc in the 
sense, that having been fed +he ~mlversally standardized inforra- 
tion on a particular restricted sulilancuaoc, it procccr!as to andlyze 
it in tile universal way and is equally ai~plicable to cach mi crcry 
restricted language. 
RESTRICTEP SUBLANGUAGES. 
Is Restricted Sublangadge Approach, RSA, applicable to all, 
or at least mostl relevant caseg 
or can it be applied only in a 
few exceptional situations?. 
It has been argued elsewhere (see 
Xaskin, 1971; Ch.4.1) that the first alternative holds true while 
in the cases in which a polythematic informational system is needed 
it seems worthwhile to treat the processed tex-ts as belonging to 
several distinct restricted sublanguages; and after distinguish-, 
ing them.with the help of a not too complicated device, t~ make 
use of the technique developed for restricted sublanguages. 
2, 
RESTRICTED SUBLAkEUAGE SOLUTIONS TO SOME PRQ9LEIiS OF HIGH 
QUALITY TRANSLATIQN 
Semant&cs and pragrna tics-a'nd. the quali ty of translation. 
Recent developments in semaqtic and syntactic theory have demon- 
strated the practically indefinite potential depth of a complete 
linguistic description which seeps to require much scarcely acces-, 
sible (at present, if not in-principle.) information on "speech act 
conditions, conversation rules, and semantic interpretation which 
must be associated in an idi'osyncratic way with the lexical item 
in question", op ':a theory af illocntionary acts", on "a theory 
of discourse which relates the use of sefitences in social and con- 
versational situations", and.on "a theory of natural logic" 
(Fillmore), while the pragmatic dimension of the text is said to 
include answers to such heteroqeneous questions as "by ~IIOK tithe 
text was produced, for which kind of audience it was meant, which 
kind of background kno~~ledge the producer of the text assumed to be 
RESTRICTED SUBLANGUAGES 
available to the audience, the time, the place, and other para- 
meters of the situation in which the text was produced. etc. 
I' 
(Bar-Hillel) . 
Now, it bs obvious that for an adequate translation, no 
matker whethex it: is liman Or.. automatic, all this, highly complex 
information must be obtained and taken into.consideration, other- 
wise the quality of translation falls down sharp1.y. It is equally 
obvious that all this is far beyond the linguist's reach at the 
present staae of linquistic knowledge. 
In order to arrive at a practical solution of this problem 
one swould impose sorre restrictions on the process of FIT. In 
other words, certiain criteria of the quality of translation shohld 
be formulated, and if necessary and possible, lowered. One might 
try to restrict the output of an MT system in khe sense that it 
should certainly not produce what the user does not actually need 
It is evident that the user of a translation of a scientific or 
techniml text will not require as much finesse and subtlety as 
the user of a translation of a literary text. So~e (e.9. Garvin) 
are prepared to go even further and construct system which would 
produce clearly inadeouate though still tolerable translations (in 
a swense nobody has even suc~~eeded in defining) in order to gain in 
speed,. Now, when "machine-aided translation" or similar approaches 
are suggested, a restriction is imposed on what the computer is 
supposed (and thought of as capable), to ao. 
The restriction on the input in RSA determines,. 01 course', 
sore restrictions on the output (but, certainly, not to the extent 
RESTRICTGD SUBLANGUAGES 
of tolerating barely acceptable tran~lations'). 
On &he other hand, 
rather on the contrary, the simplicity and easier observability of 
the material of a restricted sublanguage0mak,e automatic transla- 
tion feasible, allowing at* the same time and Cor the same reason 
for the total accountability of the sublanquage which makeg it 
possible to account for and use for the practical purposes of 
translation all the complex semantic and pragmatic in/fomation 
which might be relevant for translqtion. 
Of course, what makes it. 
possiblg is that the degree of complexity of such information in 
the restructea sublanguage is very much inferior to what might be 
observed in language as a whole. What follows, However, is that 
restricting the input of an MT syste~ to a sublanguage of a cer- 
tain type 
RSA ensures high quality translation within the sub- 
language and no further restrictions or lowering of the quality of 
translation is necessary. 
It should'be mentioned at. this point that RSA shares wsth 
"nachine-aided translation" the property of requiring a limited 
amount of predetermined and routine huvan participation prior to 
automatic processing. 
Syn'tax and semantics, lexicon and grammar. One of the major 
claims of RSA is that, 
at least in applications to restricted sub- 
languages, intricabe and labor-consuming syntactic alg~r~thms (cf. 
Me1 cuk, 1964) are redundant. 
The universal algorithm is based on 
semantics and is desicped to use linguistic information of "lower'" 
linnuistic e levels (viz. morphology and syntax) in a few exceptional 
RESTRICTED ISUBLANGUAGES 
cases of semantic'ambiguity. This emphasis an semantics rather 
than on syntax in automatic language data processing systems takes 
on a new value when compared to current discussions of the reLa-. 
tions of syntaxg and semantics in linguistic theory and the exis- 
tence of a clear-cut boundary between.them. Probably influenced 
by the tendency, at 'present prevailing statistically in theoretical 
Linguistics, to claim the priimity of semantics over syntax, and, 
moreover, to neoate 4 the existence of the bouddary, even lth'ose re- 
searchers in MTawho do not seem to be influenced by RSA also speak 
in favor of such a "semantically-based" position (e.g.. Eley) . The 
latter position is indirectly reinforcedoby the fact that purely 
syntactic contributions to the stud% (e.g. Petrick) fail to prove 
their pertinence to the problem of actual realization of MT bear- 
ing insteed on the relation of recent theoretical innovations to 
the feasibility of MT (see. below) . 
Thanks to its basie principles and internal organization RSA 
came inde~endent~y~to a justificatidn of the claim rnadd by Garvin 
khat it is aperationally more effective to delegate most of the 
gramatical irlformation used in an MT system to the lexicon rather 
than to the parsing alqorithm. 
~inguistic theor9 and feasibility of MT. RSA seems to contri- 
bute to the solotion of the major clilema concealed in this phrase 
by providing, in a.wy similar td the one discussed above in con- 
I 
nection with semantic and pragmatic problems, an interesting half- 
way position, a middle ground of a sort which in a sense combines 
RESTRICTED SUBLANGUAGES 
some rehvant properties of the two extremes-. 
In the light of quite a number of promi~ing developments and 
achievements in liriguistic theorv, - the pertinent question is whe- 
ther these have, do, or will, contribute anythirfg to ?IT, or the 
latter, as Lyons thinks, "will neither contribute very directly to, 
nor depend very direktly upon, advances in linguistic theory!' 
This basically defeatist position has at least two aspects, the 
one being that Language is claimed to be too complicated to be 
successfully subjected to autonatic processing, an opinion many 
theorists would subscribe to, and the other, proclaimed by t4T 
operationalists" (e.g. Garvin) that much of what has been recent- 
ly proposed in grammatical and semantical theory is far.too strong 
for W, and much weaker models, as a possible theoretical basis 
for practically feasible !IT are required. Ttd latter considera- 
tion is interestingly illustrated by -the fact of the recent emer- 
gence of working automatic systems of lanquage data processing, 
ouite A close, in their restrictiveness to RSA thouqh, rather contrary 
to it, not aiming at theoretical generalization, which use: "anal- 
vsis-based grammars" (cf . tlinograd, 1972) . 
-However, it i's natural for the bingu'ist to be suspicious of 
any attempts fo base an MT systen? on a theory. or a model, which 
has been dempnstrated to be inferiar to some other theory or model. 
Any serious attempt to maKe use of any lixguistic knowledge for 
any-purpose must, he might feel, be based on an adequate theoret- 
ical framework-, .otherwise the ever present danger of ad hoc de- 
cisions could hardly be avoided. What niqht 'be missed in this* 
MSTRICTED SUBLANGUAGES 
argumentation is the fact that, when dealing tzrith computerized 
applications 6f linguistics, ,we impose on thk linguistic material 
a fundamentally different phenomenon, with laws and logic of its 
own, whi"ch may be, very foreign to the nature of human language 
and the mental mechanism which uriderlies it, and this might force 
us to give up purely 1inauis.tic theories, even if they seem based 
on the properties 'nherent in man's nature, and,to adopt, in man- 
machine partnership, a cornprornlslng approach which would account 
for botri human and machine nature.. It is not unimaginable, though 
rather distressing if true, that, due to theessential difference 
betweed the two, no linguistic theory claiming .or exhibiting the 
property of adequacy to the nature or human lansuase can be direct- 
ly "computed", i,e. taken in by the computer; 
It seems,. and this is substantiated bv the material a£ some 
papers contributed to the study (e.g. Rarktunen), that the Fore 
dependent on some recent development in "PUL~ li,ngulsticst' 3 paper 
is, the less pertinent to MT it becomes. 
The contradiction Between 
linguistic theory aiming at adequacy and practfcal needs of MT and,, 
for that matter, other problems of computational lin~uistics, id 
self-evident. In this situation RSA seems to be doing a uniq~ job 
of reconciling the two extremes, since on the paterial of a restric- 
ted subLanguage it might turn out Chat the application of a .grammar 
based on adequate linguistic theory would be quite practical and 
there would not be any need to seek more feasible ad hoc solutions. 
Besides that, RSA may contribute a great deal to what is essentially 
RESTRICTED SUBLANGUAGES 
an issue between "thepry" dnd "practice" by: 
(1) providing a suitable "testing 'grcund" 'for various conflicting 
theories or models, both for those which claim linguistic adequacy 
and analysis-based ones; 
(2) 
allowing one to select the moSt.breXerable alternative on the 
basis of complete and easily accessible, evidence which night be 
relevant for the choice; 
(3) 
enabling Qne to linit the stren~th of a too powerful but valid 
theory or model bv ma'kins suitable n-odifications on the basis of 
eas2ly observable linguistic material 05 the restricted subdanguage. 
The D,aslc principles of RSA make one think of its language 
~ndependence. 
Ameri~an Journal of Computational Linguistics Microfiche 9 : 38 
LETTERS WITH VARIABLE VALUES AND THE MECHANICAL 
INFLECTION OF RUMANIAN WORDS 
Minerva Bocsa 
Universi t y of Timi~oara 
Romania 
The generation by computer of written Rumanian words faces 
two difficult problems: to produce automatically the numerous 
alternations which modify the stem and to add the inflectional 
endings, building a rich set of classes and sub~lasses. The 
mechanical morphological analysls is also complicated because of 
the stem's phonetic alternatioris. 
For example, the Rumanian words 
UNIVERSITATE -- / UNIVERSIT~~I -. (university) 
SE3lQC? -- / SERIQSI' -. / SERIQASA 
--e 
(serious) 
PUTEA -- / P_OT/PQTI ;/ POATE 
-- -- - --- 
(may 
V~DEA -- / V~/VE_EI -- -- / V~ZUI -- / VADA (to see) 
-- 
-- 
prednt the alternations 
Phonetic rules describing the occurrence of these stem 
modifications have several exceptions and must inelude the 
presence or absehce of stress, which is not marked in ordinary 
Rumanian Inf Lection 
experiments in mechanical translation from ~ngli-sh into Rumanian 
[16] and so on. Phonetic alternation in Rumanian has been 
investigated by Lombard [Ill , Felix [7] , Juglland and Edwards 
[lo] , Augerot [l'] , and others. 
The preparatory work for our automatic linguistic task has 
several stages : 
Examine the inflection of each word. 
Establish the set of phonetic alternations. 
Attach a specific variable letter to each alternation. 
In our conception [4] these are different from those of 
[9, 14, 151. 
Design a binary code for the variable letters, tailing 
into 'acc~un~the possibilities of the IRIS 50. 
Detach morphological parameters. 
Code each word. 
Punch a deck of cards. 
The card file is the Morphological Dictionary. It is exploited 
by the programs in various ways. Here the working principles of 
a program to produce the paradigm (set of inflected forms) of 
each word in the Morphological Dictionaryare presented. 
Ir. this process the computer writes the inflected forms in 
the P positions of the paradigm ? 
The stem allomorphs consti- 
tute a set A with n elements. The different distributions of the 
allomorphs of A in P are described by a see C; of grouping functions 
Rumanian Infle~tion 
spelling. Nevertheless, the words with nonc'onstant stem are too 
numerous to be considered irreqular. The method of storinq the 
several al,lomqrphs of the stem fortautomatic inflection misses 
the natural uriity of, the word. 
We* have Constructed a mechanical Morphol ogi cal Di otionary/, 
containing 2058 written Rumanian words with a synthetic repre- 
sentation of all these phonetic alternations. An algorithm based 
on this representatlon generates tne inriecrlonai noncompouna 
fbrms of these words. They, are Rumanian nouns, adjectives, and 
verbs, the main part belonging to the basic word sbck [8, 171. 
About 45 percent of them present stem alternations. 
1 
The algorithm whose logic was given in [3] is the background 
of a set of programs written in the programming language ASSIRIS 
for the French computer IRIS 50 and its Rumanian counterpart 
FELIX C-256. The proqrams were recently run at trhe Territorial 
Electronic Calculus Center of Tirnisoara, verifying the algorithm. 
The synthetic representation uses G. C. Moisills notion of 
letters with variable values [14, 151 , which V. Gut.u Romalo 
developed [9]. The setting of our research is  arcu us's theory 
of mathematical linguistics [12, 131, Diaconesculs study of word 
segmentation and the degree of regularity [5, 61 , Domonkosl s 
'1t seems that in Rumanian only 28 percent or e*en less of 
the total number of words have these phonetic alternations, but 
in our dictionary. reference is made generally to the most fre- 
quently used words, with relative frequency above 0.22% [17]. 
Rumanian Inflection 
identified bv numerals. 
Thus grouping function 00 associates 
allomorph a in A with positions 1, 
2, 5, -6, . in p, allomorph 
b in A with positions 3, 8, ... in P , etc. The different parti- 
tions of A are called allamorph configunations and syniholized by 
a/b (with n = 2), ab/c, a/bc, a/b/c, ... (with n = 3), etc. A 
varlab3e letter maps the elements of the partition into the 
Rumanian alphabet A, A, A, B, ..., 2, j3 (here pl represents the 
empty letter) . Thue the varxable letter T/C with the configura- 
tion ac/bd has the realization T in allomorphs a and c, and 
another realization C kn allomorphs b and d. Not all of the 
theoretically possible variable letters exist in Rumanian; we 
Eound 85. 
Thg set of fixed, vaxiable, and empty letters is called the 
generalized Rumanian alphabet. 
A version of i,t is given in [2]. 
Words can be represented in this alphabet in either external or 
internal code. 
The program operates in several steps which are described 
and then illustrated. 
&put. In the Morphological Dictionary, the fixed letters 
are punched in accordiance with the standard card code. Each 
variable letter is punched as a numerical prefix of one or two 
decimaA digits followed by a letter. 
Part of speech, number of 
allomorphs, word length, stem length, etc. appear as parameters. 
Rumanian Inflection 
1. Receding. The computer reads the word on the punched 
card and recodes it into an internal kdde; each letter is one 
byte. A fixed letter has zone E or F (leading four bits I110 
or 1111); variable letters have other zones. The recoding 
instruction in IRIS 50 is TRTR (translate and test). 
2. Realization. The program reads the word byte by byte. 
If the zone is E or F, it writes the byte into the allomorph 
registers. If the zone is less than E, the program constxucts 
a realizatidn for each aIlomorph and stores it in the allomorph 
register. 
The principles that #govern the decoding of a variable 
letter into realiiatiohs are given in [3]. As an example, take 
the rule for regular variable letters (zone 0, 1 . . . 7). Each 
regular variable letter has two realizations, and in the internal 
code the zone of each realization is F. The numeric of one 
realization is identical with the numeric of the regular variable 
letter, and the numeric of the other realization is greater by 1. 
The method of encodihg partitions for regular variable letters 
is explained on the next frame. 
The next program stage is on-frame 43. 
~umanian Inflection 
CONFIGURATIONS FOR REGULAR VARIABLE LETTERS 
Eight zones (0, 1, . . . , 7) en-code regular variable letters. 
Each stem has two, three, ur four allomorphs. Each partition of 
the paradigm has two members for a regular variable letter; the 
numeric of the variable letter is copied into the allomorphs of 
the. tlrst member of the partition, and -incremented by 1 inko 
those of the second member. 
Number of Allomo,rphs 
Zone 3 4 
0 ac/bd 
1 a/bcd 
2 ab/cd 
3 ac/bd 
4 ad/bc 
5 a/bcd 
6. a cdJb 
7 ab/cd 
Rumanian Inflection 
3. Receding. The program recodes the allomorphs into EBCDIC 
by another TRTR. instruction. 
4. Distribution. The proqrdm distributes the allomorphs to 
their locations in another region. The word's grouping fvncti~n 
controls the procegs. 
5. ~nflection. The program adds the inflectional endings 
to the right of the sfem allornorph in conformity with the class 
and subclass noted on the punched card. 
6. Printing, The Rrogrm condenses the empty letter arid 
prints the inflected forms. 
We illustrate concisely these ph-ases t'or two words trom our 
Morphological Dictionary, the verbs A PUTEA (may), and A VEDEA 
(to see). They have, respectively, four and five different allo- 
morphs of the stem. 
Input. The content of the card is 
PUTEA' P8UlrtA8TEh V.4 L00403 
VEDEA V9E9DEA v5 0703010 
8U, 19k, 8T, 9E, and 9D are variable letters in the external code.. 
Some morphological parameters are 
V verb; part of speech 
9 
5 
number of allomorphs 
10 
0.7 
worc! length 
04 
03 
stem length 
03 
00 
grouplng function 
Rumanian Inflection 
1. After translation into the internal cod'e' the words are 
represented in storage as 
EA 84 A9 86 F2 FO 
E6 92 93 F2 FO 
EA., F2, FO, and. E6 represent the fixed letters .PI E, A, and V. 
84, A9, 86, 92, and 93 represent the variable letters UJO, a/A, 
T/T.. E&A, and D/Z. The symbol p wtll be replaced by blank. 
2. The four or three stem letters, specified by 04 or 93 on 
the punched card, give the following four or five allomorphs. 
Thk program decodes the Irregular variable letter 84 and 
pfoduces the realizations u snd 0 (bytes F5, F6) in the allo- 
morphs a (u) and %, c, d (0) , in accordance with a translation 
table. (3) The allomorphs are translated into EBCDIC. 
4. The allornorphs are placed in new registers as specified 
by the grouping functions 03- and 00. 
Rumanian. In£ Zection 
5. The inflectional endings ale added. 
PU TEA, PU TEXtE, PO T, 'POTI, POATE, PU TEM, PU TETI, PO Tf 
-JEDEA, VEDERE , V~D, VEZI , VEDE , VEDEM , VEDETI , VXD , ... 
6. The computer condensgs the empty 'letter in A PUTEA and 
prints theinflecfpd forrps. 
The variable-letter, method has the advantage of keeping the 
Qnity of the word in the Morphological Dictionary and producing 
the inflected forms correctly. At the same time it regularizes 
the greatest part of the irregular words. The only irregular 
verbs that still remain are A AVEA (to have), A DA (to give), 
A FI (-to be) , A LUA (to take) , A STA (to stand). . The other 
so-called irregular verbs A BEA (to drink),,.A MINCA (to eat), 
A RELUA (to retake), A USCA (to dry), A VREA .(to want), and all 
the other semiregular verbs belonging td the third conjugation 
[5, 141 are regular for our algorithm, and so are. the irreqular 
nouns sod-SURORI (sister) , NORA-NURORI (daughter-in-,law) , 
OM-OAMENI (man) , etc. 
The program contains 1455 ASSIRE statements and generates 
the inflected forms for all the 2058 words included in the 
Morphological Dictionary in 1 minute 39 seconds. It represents 
an experimental verification of our algorithm and may be 
extended without essential modifications to all other Rumanian 
words, coded in the same way. 
Rumanian Ihf leetion 
Another program meant for users receives a word fxom the 
punched card without its speclal external code or grammatical 
parameters, looks far it in the Morphological Dib.tionary file 
now stored on the magnetic disk, and, if it is found, produces 
the paradigm of the mrd. Exaqples.of its outpub appear on the 
next two frames. 
Subsequent frames exhibit the complete internal and external 
codes. 
The-Variable-letter method enables us to.fom an easy algo- 
rithm for nnrphological analysis, as indicated in [2] . 
Rumanian Inflection 
TRANSCRIBED OUTPUT' 
CuvTntul cexut : PUTEA 
~~s~unsul ordinatorului : 
1. PAMDIGMA VBRBULUI A PUTEA 
Nr.prs. Prezent Imper,fect Perfect Mai mult ca Prezent ImQe- 
indicativ simplu perfect conjunctiv ..rativ 
Sg. I POT PUTEAM PUTUT PUTUSEM POT 
11 'POTI PUTEAI PUTUS1 PUTU'SE$I POTI POT1 
111 POATE PUTEA PUTU PUTUSE POATE 
P1. I PUTEM PUTEAM PUTU~ PUTUSE& PUTEM 
I POT PUTEAU PUTU~ PUTUSE$ POATE 
Modurile 
ncpersonale : Infinitiv PUTEA PUTERE 
Participiu PUTUT 
Gerunziu PUT~ND 
Rumanian Inflection 
TRANSCRIBED 'OUTPUT 
~uvf ntul cerut : VEDEA 
~&~unsul ordinatorulus : 
1. PARADIGMA VERBULUI A VEDEA 
Nr. pers.. Prezent 
Imperfect Perfect Mai mult ca Prezent .. . Impe- 
ind5cati.v sim~.lu perfect con junctiv rativ 
,111 VEGE 
- -. 
P1, I VEDEM 
VEDE'AL VAIZU$I V~(ZUSE$I VEZI 
VEDEA 
u .) 
VEDEAU vlizud V~ZUSEG VADA 
Moclhrile nepersonale : Infinitiv: VEDEA VEDERE 
VEZI 
- A 
Part-i'cipiu: VAZUT Gerunzi~ : VAZIND 
Rumanian Inflection 
GENERALIZED RUMANYAN ALPHABET (EXTERNKL AND INTERNAL CODE) 
Zone 
Num . 
. 
0 
OnOO,fl/A 
1 
0001 
2 
-0010 
0' 
OOtlO 
--5 .-. -- ,. 
$A 
' a; 
%;/A, 
I"--- 
1 
0001 
2 
0010 
3 
0011 
4 
1A 
Afi 
. - -h l--... 
1E 
E/0 -- 
gf 
a/ 
3 
0011 
5 
0101 
----- 
4 
01.00 
2: 1 3: 
?/A / 11: 
a/b ; 
Q 'T 
I 
. 
I - - - - ----- b 
5 5U R.. U 
t 
U/O 0,'U 
, 
4: 
?/I 
6 
0110 
7 
0111 
8 
1000 
9 
l~Ol~/C 
A 
1010 
t 
B 
1011 
C 
1106 
D. 
1lQLa 
E 
I: 1.10 
.p - 
1111 
I. 
I 
- 
I 
6. 
0110 
-..- .. 
5A 
;/A 
.. : .5i. 
.EJA 
.-- - 
4A. 
A/; 
4i 
' E 
Y 
M 
. . 
I 
I 
7 E 
0111 
I 
a 
I 
10 
I 0/0 : 
F 
7A ii 
~/i 
3s 
S/S 
3~ 
T/C 
3 T* 
T/P 
• 
, 
-2 
I 
PS i 
p/S i 
A 
a 
- 
4s 
s& 
4C 
C/T 
4T 
T/T; 
4D 
D/Z , 
42 
z/J 
-1-1 
I'u 
6i-1 7A F A 
a 
PC i J-C 2C 
!C/p !C/g 
X/E; i/~ 
K 
5s 
S/S 
5~ 
TIC 
5T 
?/IT 
5D 
.Z~D . 
I 
E 
L 
I 
.2 G 
G/I 
2N 
N/B. 
2B 
B/P : 
.-.- .- 
I 
I 
I 
IG 
I G/la 
.. ,.jify,/ 1N 
a/~ ' 
I 
$/H 
. 
L 
I 
6s : 7s W . 
s: 
'.I 
s. 
'~1 
* 1L 
I L/% 
S/S :.S/S . w', I 
I 
i 
xi 5 
. . 
I 
1 I 
L 
6C ' 7C 
C/T ,i'C/T . . 
YI 
! 
T, 
b 
T 
D 
' Z 
,,,, J 
6T 7T P 
T/T' ] T/T 
I 
1 
C 
6D 
,D/.z 
I a 
.LID N 
D/Z Y, 
62 H 
Z/J ; 
- 
. - 
.B 
b * 
. . 
Rumanian Inflectipp 
GENERALIZED RUMANIAN ALPHABET (EXTERNAL AND.INTERNAC CODE) 
E 
1110 
A 
*A 
I? 
1111 
A 
7 
, B 
,1011 
C 
1100 
7 
0111 
- 
8 
1000 
9 
1001 
A 
sol0 
B 
1011 
C 
I100 
D. 
1101 
E 
1110 
1 
. D 
1101. 
A 
1010 
+- - 
9 
100i 
Zone 
N urn . 
F! 
A 
%OA 
A/E/A , 
a/bd/c 
8 
1000 
K 
W S 
I 
I 
---_---. 
. j 
' X s 
L 
i 
I 
-. . -- 
I 
-- , 
8 1 
L Y C 
1 I 
8D 
D/z/~ 
.. adb/cd 
8L 
LI-p % 
acjbd 
8C 
C/P 
ab/5cd 
9A 
.A/% 
ae /hc.d 
0. 
0000 
6- 
E 
92 17% 
T/T/% X~A 
ab/c/de ---- -.- - abicd . --- - 
9s' 18A 
si; PA 
abd/c 
.- -- abdlc 
9G 19-A 
cia. %/A 
ac/bd - ab.c/d 
#A 
-A/ja 
ad/bc 
--I I. 
----- --- -- - --L----L--.+ -- 
- 5 
P T 
i 
- 
I 
i 
I 
G T 
# 
9A- 
A,& 
abe/cdf 
933 
1 
0601 
2 
ll~l 
M 
- 
8G 
G/P/~ 
a/b/cd 
8E 
EJA 
ac/bd 
12A 
h 
I 
N D- 
l 
--- 
H z 
I 
0010 
3 
o 0.1 r 
4 
0100 
. 
5- 
010r 
6. 
0110 
- 
E/I 
a/bc 
80 
obu 
abc/d 
8U 
u/o 
a1bc.d 
85 
s/g 
abc/d- 
- 8T 
T/T 
abd/c 
E/X/A +A'/E/~ 
ac/bd/e -/ S/b/cd 
ga 13~ 
D/Z ! ?/(AH 
abe/cd a;/b/cd 
' -. -r-.-.-.m.-r.--. " 
92 14~ 
. D/Z/g EJA 
, ab/c/de ab/c'd . 
15A 
9J A 
~lz/(a - ' I,&A 
'ac/bf /de ab/c/d- 
9T • v i6~ 
T/TL% A/A 
a;;/b/cd ad/bc 
I 
I 
I 
B J 
. - .- 
' 
. 
Q 
- . r, , ,,. --. , _A ----.. -L 
, 
1 
I 
I 
I 
! 
I 
RI U 
.V 0 
I I 
i 
I 
L* 3'1- 
Rumanian Inflection 
American Journal of Computational Linruistics 
Microfiche 9 : 5 3 
SECRETARY-TREASURER 
A. Hood koberts 
As of March 22, 1975, ACL had received dues foi- 1974 from 
individuals 
libraries 
Eoy 197'3, the comparable figure is about 400. 
FI NANC I AL ACCOUNTS 
Cash on hand August 1, 1973 
14ember§hip* dues $3,886.00 
1BM contribution for meeting 300.00 
AFIPS distribution #of surplus 3,228.00 
Receipts at 1933 annual meeting 705.00 
The Finite String. 
Office supp1ie.s & expenses 
Expenses at 1973 annual meeting 
Expenses. for 1974 annual meeting 
AFTPS dues 
Bank fee 
Certificates of deposit 
Cash on hand July 25, 1974 
Savings accounts 
Net worth July 25, 1974 
American Journal of Computational Linguistics 
Miorofiche 9 : 54 
EDITOR S REPORT 
The 1974 series of this new journal has been produced from the 
first 44 manuscripts received. 
Of this number, 
10 have been published as microfiches 
4 have been published as notes in The Finite String 
10 are with the author for minor revision 
7 are with the author for major revision 
9 have been rejected 
4 were returned to the author withoot substantive 
comment because tKey fall outo'ide our scope 
More than 2 out of 3 manuscripts go back to the author for some 
reason. Perhaps somewhat more than 1 out of 3 will eventually 
appear in this Journal. If our rejection rate remains low, 
it 
will be thanks to the microfiche format, which permits us to 
publish more pages per year than most journals. 
The 1974 series contains about 900 frames, As more contributors 
take advantage of the space available to them, individual fiche's 
will be filled more often, As nore workers in the field become 
aware of the quality of the audience, the rapidity of publication, 
and the value of space in which to make a full presentation of 
methods and results, the number of contributions per year will 
increase. We anticipate a larger series for 1975. 
A new policy, effective immediately will probably make it pos- 
sible in almost all instances to give the contributor the answer 
in &out a month that has been the intent of the editor from the 
beginning. 
The editor acknowledges with gratitude the support of the ~ationai 
Science Foundation, the general help of the Center for Applied 
Linguistics, the support provided by the Executive Committee and 
membership of the Association for Computational Linguistics, and 
the imagination and effort rendered by several persons--not in- 
cluding the editor--which led to the creation of AJCL. 
The work 
of the Editorial Board has been, and must continue to be, one of 
the foundations of the journal. 
Those whose names appear with 
mine on the first frame of this fiche work hafd and well to make 
the journal useful to its readers. 
The comments of all concerned are always welcome. 

REFERENCES 

Gorodeckl], B. Ju., andwV. V. Raskin, 1971. Nethods of Semanticw 
Investigation of a Restricted Sublhanguage, MOSCOW: MOSCOW 
University Bress (in Russian) 

rJelfcuk, I. A*, 1964. Automatic Syntactic Analysis, Novosibirsk 
USSR Academy of Sciences Press (in ~ussian) 

Raskin, V. V., 1971. Tgvards a Theory of Linguistic Subsystems, 
MoscoQ: Moscow University Press '(in RUSS~~~ - an English 
translation is in preparation :by Mouton) 

L'ehrnanh., W. P., and B. Stachowitg, 1971
Feasibility Study on 
Fully Automat~c High Quality Txanslation , Griffiss Air Force 
Base, Rome Air 'Development Center, RADC-TR-71-295 

:Jinogr,ad, T., 1972
Understanding Natural Language, New York: 
Academic Press 

1. Augerot, James ED A study of RunaniiAn morph'ophonology. 
Dissertation, university of Washington, 1968. 

2. Bocsa, Minerva. Coda.ge de 1 ' alphabet g&n6ralis& du Roumain 
pour l'ordinateur IRIS 50 (FELIX C-256) . Cahiers lde lin- 
guistigue the/orique et appliquge, NO. 10, Fasc, 2, 1973. 

3. Boc~a, Minerva. Alqoritm pentru generarea cuvintelor limbii 
romane in ASSIRIS. Revista de analiza' numerics si teoria 
approximafiei, vol. 3, fasc. 1, Cluj, 1974. 

4. Bocqa, Minerva. Dictionar morfol ogic automast a1 limbii 
romane pentru prdinatorul IRIS 50 (FELIX C-256). Disserta- 
tion, Universitatea din Timi~bara, 1974. 

5. Diaconescu, Paula.. Contributi) la definirea-~i clasificarea 
verbeior regulate in limba r@^ana'. Studii si ~ercet%ri ling- 
visitice, no. 2, ah XI, Bucureqti, 1960. 

6. Diaconescu, Paula. 
structura* ~i evolutie ?n morfologia s&- 
stanti vului rom8nesq. Kditura Academiei ~e'publicii Social- 
iste ~omsnia, Bucure~ti-, 1970. 

7. Felix, ~iii. 
Asupra alternanfelor fonologice din f lexi,unea 
verbala romaneasc*a. studii lgi cercetFzi iingvistice, no. 6&, 
an XVI. Bucureqti, '1965. 

8. Graur, Alexandru. Incescare asupra fondul ui principal de 
cuvinte. Bucure$t$, 1954. 

9. Gufu Romalo, Valeria. Morfoldgie structural: a lirnbii 
roma^ne. Editura Academiei ~epublicii Socialiste ~omznia, 
Bucuregti, 1968. 

10. Juilland, Alphonse, and, P. M. Ha Edw,ards., The Rarnania~ verb 
system. Mo~tm, The Hague, 1971. 

11. Lombard, Alf. fe verbe roumain. Lund, 1954. 

12. Marcus, Solbmon. ~in~vis ti csm t ema.t,i cz .- 
Editura didacriticg 
$i pedagogic!$, Bacuresti , 1966. 
itumani an Inflection 

13. Marcus, Solomon, .dlgebcbic linguistics, Anal yti tal models. 
Academic Press, New 'York and London, 1967. 

14. Moisil, Gripare C. Probleme puse de traducerea automata, 
Cohjugarea verbelor in liqtba romana scrisa. Studii $i cer- 
cetsri lingvis-tice, no. 1, Bucureqti, 1960. 

15. fioisil, Grigoxe C. ~robl8mes posgs par la traduection auto- 
hatique. La decllnaison en rbumain &it. C-ahier de l'in-- 
guistigue thegrigue et appliqu~e, no. 1, Bucureqti, 1962. 

16. Nistor Domonkos, Erica. dlgoritm de fraducere automat: din 
limba enoldza in limbar romana. Editura ~idacticz $i Peda- 
goglca, Bucure$ti,- 1966. 

17. Sgteu, VaTeriu. Ohservapii asupra f recven$ei cuvintelor in 
sperele Unor scri-itori ramani. S tudii si Cercetari Lingyi s- 
tice no.. 3, Bucure~ti, 1959. 
