File Information
File: 05-lr/acl_arc_1_sum/cleansed_text/xml_by_section/metho/79/j79-1081_metho.xml
Size: 128,653 bytes
Last Modified: 2025-10-06 14:11:15
<?xml version="1.0" standalone="yes"?> <Paper uid="J79-1081"> <Title>CHARACTER.PAIRS. AND CHARACTER-TRIPLETS</Title> <Section position="1" start_page="0" end_page="0" type="metho"> <SectionTitle> THE FINITE STRING </SectionTitle> <Paragraph position="0"> Released-for publ~cation March 24, 2979 With this issue, David O. Hays completes his term as Editor @-f AJCL snd breathes a sigh of relief. Personal matterg have made the last two issues of AJCL for 1978 excessively late. The next issues of AJCL -wSll appear on paper; but the circumstances of the mumepa nuggest that digital magnetic recording and direct wire transmission ill be suitable for experimental use shortly.</Paragraph> </Section> <Section position="2" start_page="0" end_page="0" type="metho"> <SectionTitle> AMERICAN JOURNAL OF </SectionTitle> <Paragraph position="0"> in Mana<7ement,, a subgroup on natural Lang~,a.g~ processing visited the soviet Union EUcon: R.ay 25 throug'r Jnn? 11, 1978. The qr3up azt with szient ists i:~ Noszow, N~vosibir~sk. Leningrad, 3nrlKiev. There were fornral rn2e;tinqts and pre%sentations of technical materia I, 3nd 3 lso m3ny inFornd1 discussions- This report pres=g t5 a view of Soviet c~mput 3ti2,nn 1 linqui sti:.s wh i ch emerged f~om these 3is~ussi~ns.</Paragraph> </Section> <Section position="3" start_page="0" end_page="1" type="metho"> <SectionTitle> Tbc~~llmS./rlSSP Science Exchange on ~pplisations of Comput?rs </SectionTitle> <Paragraph position="0"> to tlanaqenzr t in=lu?es many sub-ta.;ks. The exchange in natural languaq~. prxess in7 is one task undet- thq topic &quot;theorst iza 1 f oun dat i on.; f~r softwarir in applications in, rconomics and man~rle?nent~*, The 2xchanqe in natural languqgc. procsssisq uss to hvo bequn in June 1977. However, a scbelulei tri.p by 11-5.</Paragraph> <Paragraph position="1"> qrirhntists WCI,? cavzslL~3 at the fast minute by the USSR si3e: t9e . F, 'JrSr Ch2irrncrn of the I1.S./USSR Joint working Croup 3e Scicntif LC an1 T~zhn,i.=3l Coopratio.\ in th? AppLicat ion 3f Conputerr to * anajernnnt ; Sue Rogner, 4. Em ; Jayco Prieirna~, ?rpart\ent of Zoapufer 3114 :onttnunica tion Scienczs, The Universit p. of qichigin; tobn H3) houl, Bolt Beranet and Newman, Inc., Cnmbridqe; 5t3nl~y Pctrick, Mathematiss Dspqrtment, f3 r T. 1. aa ts3n R~searr: h Lontze, Yort town ilri3t.t~: Saliy Sedelou, D?.qartmc.ntr, of U?gu~+sti~= and Co~puter Srience, Oniversity 3f Sansas: an1 Udlt'.~r A. Seilelow, Departments af Fociolngy 31d computer Scien?e, UR $v~rsfty ~f Kansas. The 0-3. lelega tinn vis accampaniel throughm~t the trip by A. 5. ~ar%n'p@ni 3f No~osibi rsk .</Paragraph> <Paragraph position="2"> This report gr3up5 tog~ther similar work dane in differe~t locations- rllh3: main patterns of the natura hnguage pracessi~g and theorem-provin! systems cam he view=d 35 based an (1) linquiqt ~CS, (2) ar t ificial intelligence, 9r (3) lagibc, a lthoi*~ h tho distinctions ara tr, somt. extent arbitrary. We also in over~i~w oC tha compu tsrs and yroqramming 13ncjuages available f st w3 t iA camputlt fonal linguist~cs. Work on Iexicography, thesauri, and speech re?ognf.tibon uas also discuspeil on the visik, halt is not zavct~?3 SD this report.</Paragraph> <Paragraph position="3"> The m3i.n root.; of ths Iin3uL(stically-bas~d work are the mean &ng-.text mod91 of Fel'chuk, d~pendpnry grammar, 37d ttansEUormationaR qr%mmqr, They are variausly interpreted by diff er~nt systeas.</Paragraph> <Paragraph position="4"> Zopa ShL yayitqs, taboratory of nach ine Tch its la t io3 Tnstitute + a?rexan Lanljtia3es, descfibe3 an English to Russlin mach ine translst ion system under development sinze 1972 and bss?d primarily on +he mesninl-text nroAoL. The rzpresent~tion is a &ependency trea, with word order informtion, n:,rphalogy awl senantic/s~rntactic valr3ncies. This structure preser9.s all tae turf ace data but is also close t a semantic ceprzsantation ~f ths text. f her? i3 a Aictianary and a gramaar for aach Languiw-.. The* aramrmr rrll~t ar- of the two forms: &f <structure> t,b,zg, <c?ndition>, an3 iC Cstacture? thgg <traasform%tion>. Semantic: inf~armat ion incluies senant ic descr ipt~ons of l&rical 2nd rn3rph013qic31 pn its and the semantic accept3biLity af word pairs. There is a dictioniry OF 30,700 lexsmes, described in terms of 30 semantic pcinst ivt-s The syntactic arid rzemfntiz dtructutes are coqpatihlo, so ansl'ysis q3es onl'y a3 deep as is nates-ry for a qbivu& s~>nhnce. Shalyapiha's qroup worKs Dn Lrncruist ir! aspecr s dnl y; ther- is go ~a~prlter impleme'ntation.</Paragraph> <Paragraph position="5"> Uri Aqresyan 3tso wares with the msani'gg-text ~odel and wi.thmackrne trlnsl~tion as the~jo3f- His work is nri-rltgrlly on Prenzh t3 Russian trhnsLitii>ns, but he aLs9 works .or1 ~nglish. His wnql ish grlmmar is cia id to b he most complet~ evrr puhlishel : thh nossiin qrImmlc will soar1 appear. Th- limjuistic adel gill have fovr v3 r ts: m:,r ~holagy, deep syrlta x, sdrf a::e synta x, a~d r,om-ntics; bowever, the zurren t reduced no4~1 l~ks seiuanties. A qirtion3ry giv2s; fgr each ~r3 its morphol~gy, its syntactic aud seqa ntik f-%+ur=s (there are 150-syntadt ic features; 500 semantic* features), the semsntic: criteria for possible govlernin~j word.;, an1 selccti?ndl q~strirtions. Rule sch'ema or wsynt3gmas'' go frrrn morpheme structr~re to a surface syntactis structure that is 'in unor 3erod' 4znendency tree. The-re arc a hottt 2) 3 syntl pas f ~r Russian, each rppresentinl 29 rules. A syntlyma allbus a trze with X over Y ts be constttuzted from a string c~ntaining X and Y un lor variaus complex, conditions. The. Irxica 1 infornafion and th.3 synt a-jmls determine tho transfotrnation from worf sttih r to surf ace-syrrtactic structure. A deep structure is then define? ay ltp-arap_hra~ti~'t I. whish convert. for exsm~Le, s&g&& to del.'kxgr when tha ob-ject is 3 -- blow. The daep structure is no lon~er lanquaqe-spcific but is r1nivers31. and serves as the hasis for translhtian hetwe~n languages, 4~resy3n Stre~qed t~e valde of cnntinuing to work on the same linpistic moilel in or3zr tFi complete its dev;lopmant; he eontrAsted this with the attTtq1e ok soinr current Aa~c icap lingu isrs.</Paragraph> <Paragraph position="6"> The lin~ui-su&quot; Iakalev, df thq Sconomizs Iz~stittrt~ L ?i developi ny 3 nat ura1 lwicjua~e interface for (r ?1 ta basc sypd-en.. Tbjs work has cgrnput~~ suppott an?, is runni,nn soon *in 3 Isrqe fact~ry. Th2 8aturaL lanquaqt. srlt~se+ has sentencss sucla 35 'lwh?t 1s the ntrinher of wqrkers aEU <rype> in <pl&quot;ac~>~' an4 is said to h- easy %r ocon~mis~:; to 3earri. The systm is based 3h v~ry recent modslc 3f t ransf or~atronal gramnar: Isk31ev ment ionzd &quot;traccls&quot; 3 sole 3f Jacken33f f*s theoric*~. 'Ih= syst~m TOPS Er3a input to ti deep ~>tr~lctt~re f rqmv which it constr:~zts a farmula i3~ tb~ co~r~t~t ion of % tl~tnericaL result.</Paragraph> <Paragraph position="7"> AT-hated syst~ras are beinq developc3 at th Colnpi~ting Centxof, the Rcai-my of Sziences nt MOSCOW, 11niiar tt,a direction 3f Vi -+fir Driq brin snil at the Computing renter dP the Siberiir, Pivisiort ~f th~ Rzideag of Science.;, Nav33j trirsk un8er t12 airection r~~ ll?xan3e-r Narin'yani, in ~rshdv's .group.</Paragraph> <Paragraph position="8"> The system ?enonstrated to us in Morc~w wls DrLOS (Dialqqie Inform.itio? Loyical System). This work is heaviiy influehced 3y artificial intelligc~nce work in the U. S. 18ri.brln1 spent scvan monthsat Lf-F., w3rking uith William Rartjn and with Carl Hewitt-) DTLOC; is written in LISP and runs 3n the BESrl-6 co~~pli+il.c in Ho'scow, a well as Dn a PDP-11/45 at the Intczrnatianil 'Institute for Appli.23 Systems Analysis in Laxcnburg, Austri.3, Ths sptern is intenled both to test various apptoaches to natt~rll language processi.nj and for praet ical a ppli.~ations., It zonta3.q~ an hTN lin~uistic processor and a semantis prxessor ba-sea >n f rsnes. The cut reat applications are3 i airline ti.ck? t reservations: the 3?morjstration was however on 3 very small data haw of ATB Ifat trraf Languag e Systems (includin? DILOS, qU5, RFL, Out, anrl LlfNAP). The systen was a He to answer simple natursl lanquaqe questions from the data Ddse hut it was not possibLe from the demonstration to clzt a good feeli~g for th2 actual rmfe .oI 1 anou age aesgptei.</Paragraph> <Paragraph position="9"> rJarin*yanies ?coup in Novosibirsk has 17 pmpla, inzlu3incf 6 linyuists and 9 mathem3tgcia n~ and programmer=., Until 3 f?w years ago, the worlr followel Mel@Chukes morl3L. Phis has nou besn abandoned here aqd work proceeds along faur lines, so far ralativ~?Iv inde~en.1 zntly: (1) Marine yitni is dewelping a fggcxrl Linquis$ic moig& which zombines depen3?ncy an3 constituent --I -structure in n mi, xed multi-level reprs~enta tion. Analysqs proceeds by loc31 m2dification of the qraph structures, axparldilq and compr~ssing z3se frames at var ousi.1~~els. l'he li.nguisti~ mo3el so far inzlu3rzs form1 descri ptiqr~ 3f aiverb groqps 31d adjective .jroups, This formal model has now bezn written up, but so far is not implemented. (L) The semantic guestion-answmi~c~ system VOSTJK-O contains a formal model of time. 04 the basis 3f texhs of sentanyes such as wProm the 3rd up to the 13th of finrlh flike was in ~oscow~ it answars guestions'like Vhere uas Yike at Roon on the 17th oc March?&quot;, The system is coq33 iq SETL 4nO was ileaonstrate3 t~ US. While the natuta& language frqment is still smakl, even For a m:,de'l of time, (el g- nd ti&quot;me sdvsrbials), t,+e inf erenciny schame workad suc~ssEUu~lY. (3) Sev3y3 1 t~applicati:,na1'~systemcs are being devel ped- The first of: thp-3, the PL-1 fBrnini&quot; or &quot;toyw system Z4PSIR20 use': assenti,aIly QO syntactic 3r,atysis (thmqh i.t relies heavily on word order). It has. a well-3efined sunject domain, a dats base 3 oersonn~l inf~rmation, and 3 vswers questions such 3s *'who nnde~ 30 earls more than d~eraqe?'~ (Salary information is ~uhlkc in the US5R. L Tn this vzry Limited suhI~ct dorn;li.n, th2 anwoazh works uclL, The t'midlitt applicationnb system is unbr d~velopment and ins morp synt actically oriel t ,.dl 1 t will cqntain a nondetermi.nisti.c bottom-up p3rs-r FSc a big pry context-sensitive gramqar wit 11 8iscontinu~u~ ~onstituef-ts. (4) The final subjtonp is tbc proqrammin~ languap group; it hss .irn~lem\ented 3ETC on the BESYfi * In fl~scow, at PTNITI, the ringuist E, L3. Paducheva and the mathemat iciin I!. D. Rorelska ya ace developin7 j?intl y an anproszI1 to natural lanquaqf ~ramssi nc, base4 Dn both t rfnsfarqat i7,r.tl grammar a13 first-order 0. The cl~rrpnt 3onai.n is converv theorems in ge.ornetry. The system is able to prqzess- ~eometcy theorems and prbaace their 19conv&rse theorems&quot;. In this systzm the semantic repr~sentatian 1anquage is first warder loji.2.</Paragraph> <Paragraph position="10"> Algorithmi.3 procedures for analyeis and synthesis have been 'deoelop~d, ss we 11 3 s processing procedures within the logiz.</Paragraph> <Paragraph position="11"> The linqnistic psct of the methoit is based on trabsfarmationnl grammar.- As i th. case with most of the Soyiet work 3x1 transEUc)rrmational grammar, the deep struzturs uses dependenzy grammar t~~hec th%n constituent st rlict ure gr 3 rnmar. T$ e transformations ar9 originally written in the Foruhrd djxectio~, i-em from jeep to surface structure. Analysis is 33ae using a wrsverse?'8 vecs inn of each ttaasforrnati~n (not ohtain-ja ~~~thmatScatly). While t be f orwarJ transf~raatiaks ar e libi+pe~l~ent of oc let the reversal rules are strictly orler~l, for efficjeney. There are 30-37 transforaabions, each eipress?3 as a structural description, giver as a tenplat?, and a structursl chanye, qiv~n as a ssquenct2 of el-n~ntary operatiom.</Paragraph> <Paragraph position="12"> a The work I s i 3 vc.1ope.I in detail, but has no comput2r implem~ntation. The systtm is said to co?te~n interestiiq sc,Lntions 29 pt:,hLsms of qt~antific?tion, neqatlon, a%d con j unct iov re\l uct ion. Tm e authors raport 32, wit5 son tr amusement, tbar tPe description of the work was printed in 42,030 copi 9s.</Paragraph> <Paragraph position="13"> The current work at the University 3f Leninqrad u~der 3 Tsaitin, Facolt y of Enqine~ring and irathematics, was clsscribed to us by others as based 3n 197ic8 bat l'seiti-n himself took a phi1osopCi::~l appr3acb jn his discussions vit~ us. His renarca vere mor- sagge-; tivl thcEf 3zscripti.v~. He iniicated that his approach t> catur11 lanquage was bf analogy to programming 1anquast3s8 usin7 mlz ros as: in operatinq systzas. Hz cLar'lPad &quot;that thete is n3 such thing as meaningn, but sail th3t Itis approach di3 us% pr~sedural semantics. His pcevious work 3'1 complexity and ti aor~n-proving i.s not related to his wark 3n natural languaq~. However, heid P argue 'that a natural 1 angu'iga systea for computars should teflect the fact that natur31 langua )e pscf ormanca by people doe? not r;epu.ire exp3nen ti.al tima, Tsertln's nun current u3rk is no& on natural 1an3uage, as he is busy vri,ting 3 9LGQL68 i.mplerr\entation.</Paragraph> <Paragraph position="14"> Tsc rti n an3 t iak.ina, formerly of the Fad Lty of Phi lolog\, also talke? dhout ssveral esrli.er natural lanau~ge spst~ns ~hi.>h I am unclble &a 3lst1nquzsh. They are descrrbed i.n a number 3f oubli-cation; from 1366 on. In gererat, they vrpL~y Sepandrn- y grammarc, 3 use tran.;f~rma bions d11ri.n~ ~yntd2ti.~ dnaly~i.3 -Ppst~i.ctLonz on tha gramat are stated :in the pcedizste calculus and resolution 'th9or-m-provinq i,s used. -'The goal is Enqlish to Rus- ran translat i.on of scien ti.Ei.~ tlrose-Thc system of 3. Kapito?ova, riead of the Laboratory 3f A~pli.e;i Cybei.neti.cs at. thz Institute of Cybernetiss at Kiev, i.~ an in+~rac+i.ve th~.,r~rll-~rovi,n'~ system fbr mathernnti.c~l text=. The 0hyerti.v~ is to be able to fi.1 l i.n tha sta'hdard uapq in proofs, as indi.cata3 by &quot;it is obvious thatw or #'as irk the pro:,f of tho nrrui~us Theorem&quot;, The text i.s fitst proczsse3 manually rnto a hi$h'Ly stylrzed mathanst1ca1 languaqe- Only the form 1 material, theorvus rnil pro~f s, is analyzed: diszussion i.s tres+~C as comment and i..; ianored by the programs- SeveraL larae texts, including Curers an3 R~rner Al9ebrai.r Thyzy of Ztou_~_s, have hosn prep rores5~3.. T~P t heorem-prover i.~ \tailore& to the specific mathematicd. iom?rn- Xt uses resolution thzorem-pravinj, heur i.sti.c technr ques, as well as speci-a1 mathems ticil and logi- a31 tachni~ries. TM system has been programmed and is about to ~e tried out o a recent thesis.. This projest is of ton yescs .lur~tion, and has had a minimr~m or 12 people.</Paragraph> <Paragraph position="15"> Tnter?(st in Man tscfne grqmmaru was considersbl?. Fly talk in Moscow w3 s very we11 attended, arid there ware mnrlv gosd qttestions. i'he audience was qeneraI.11 famiILar with Rpntagu~~s w.o r k- an3 with rez~nt papers or, the tapiz in ~rt ieicilr. ---- 1 ~ntelligenz- qn?. ~~IIOCP tical k&i.ggi~tics. The interest seeme3 to ------- ---cqe fr-nn P morei g neral interest in bgic as a kr,owleZje yepr~s~n t3t inn i'n nata~a 1 langudge -systems. Aqafanov in NVsiis is 3 Is3 intmste3 iri tb,e possible applicntions ~f Yont 39~~ gra mnr to pro~ramm in7 languaqes.</Paragraph> <Paragraph position="16"> Coaput?r scc~ss appelrz 6a be mucF more Jiffizult to obtain for caaputft ion3 1 linq~l&s*s in the Soviet Union. flany of t're pr,a-jects h3if no cmputsr support, even though they wsra in ares where cnmpAt~r testing of grammars or the~ries cauld bs very ussful. Y 3st of the compllting was on the second-generatian' co'0puYer 9ESfi-6, a1 t bough them axe more rezent computers, e. g. , thc ES-ED4 (Qyad), series, availabie- for othzr porposes. 0.3, comp~~ters wers or orrler from Hewlett-Packard-, CDC, aqd 9urrouqhs-The termi~als 3 sau were mainly graphics terminaks from Eastern Europe, with both Roman and Cyrillic character sets2 and seortled fine in use.</Paragraph> <Paragraph position="17"> There is much interest in advanced progr3mming languages.</Paragraph> <Paragraph position="18"> SETI, is imp1 ernentef in Rovosibirsk. (This is with the akd of the U. 5. /USSR Science exchange. 1 'In Moscow, PASCAL is imp~emantal. Tn I.eningrrl, Tseibio is impie'menting ALGOL68 f3r the Ryad seri?s of compllters, compstible with the IBPI 36% We di3 have occasion ta see some interaztiva systems in opet ation, ~h; Id ngtiages were impressive, but the ptogrammx support was not, T'h~zo spame5 to be few error Ziagnastics- Wh?n th-re wore zr8sh~s it was not possible .to tell vhizh were due to th? computer aarl which' to tbe programs, ark onns tur31 langulje processinq in tha USSF seems to be alonq three m'ajnr lines, The work by lcinguists is motivated by m3ch ine translati~n. Tt relies on VPLS~ORS 3f Mel lthtikqs me2ninq-te:<t moclel, witb. some type of tran~focmations on a flz~~nd-ncy base. Tt is characterized by 2 great deal .3f soph isti cated devn3 :,pinent of large grammars, by large groups >f linqrrist-'s, ~br~t i5 without- computer support. The artifizisl inte llige~ze work is dir?cted taward d3ta base inPSotmat.i~n systems, is at sn ed-tliec state of devel3pme1t. 386 is heavily has9d on U.3. work. It is caaried out in Zomplitinj Centers aa? has good proyr3rn~ing an? computer suppart. Th loqic-based wock in carrie? out by_ individuals or small groups in sever31 1.ocations withaut rornputsp swport, and by ona lacge group with colnp u tctrrs,</Paragraph> </Section> <Section position="4" start_page="1" end_page="1" type="metho"> <SectionTitle> CONTENTS MACHINE -%TRANSLATION OF CHINESE MATHEMATICAL ARTLCLES ......................... S.-'C. ,&oh. L . Kong, and H.-,$. Hung THE CHINESE UNIVERSITY LANGUAGE TWSfATOR (CULT) . A REPORT .................................... P.H. Nancatrow AUTOMATIC SCANSION OF SANSKRIT POETRY FOR AUTHORSHIP CRITERIA .................................... D . Wujastyk THE MIzAR-QC/~OOO LOGIC INFORMATION LANGUAGE ............. Ama Trybulec THE DISCOVERY OF SYNTAGMA'EIC AND PARADIGMATIC CLASSES ......r......*.~.....oomoo~..m~o .................. J.G. WolfPS REPORT ON A COURSE ON THE USE OF COMPUTERS IN TEXTUAL ANALYSIS AND BIBLIOGRXP~Y HELD IN THE COMPUTER UNIT OF THE UNIVERSITY COLLEGC OF WALES </SectionTitle> <Paragraph position="0"> @ERY$TWYTH. 20-14 APRIL 1978 ..................... P . Sims-Williams</Paragraph> </Section> <Section position="5" start_page="1" end_page="1" type="metho"> <SectionTitle> COMPMBLE COMPUTER LANGUAGES FOR LINGU LSTIC AKI) LITERARY DATA PROCESSING: PERFORYtlNCE .................... .. . M . Boot </SectionTitle> <Paragraph position="0"> A PARTIAL-PARSING ALGORITHM . FOK NATURAL LXhLti; 4t:E ................ TEXT USING A SIMPLE GRAMMAR FOR ARCUYENT5 Pi J . Sal lis ................ WANTED . COMPUTER READABLE DICTlONAKZES T.D. Crawford</Paragraph> </Section> <Section position="6" start_page="1" end_page="1" type="metho"> <SectionTitle> COLLOQUIUM ON THE USE OF COFIPUTERS Ih TEX'TUriL CRITICISM: A REPORT .................... z........b....... S.Y. Hockey </SectionTitle> <Paragraph position="0"/> </Section> <Section position="7" start_page="1" end_page="1" type="metho"> <SectionTitle> THE WORDS OCCURRING IN ENGLISH IDIOMS ........................... D.J. Wright PROPOSED CRITERIA FOR PUBLSSHING STATISTICAL MSULTS ............................... D . Ross and B . Brainerd ON THE TEACPING OF RUSSIAN NUMERALS BY USING ONLINE COMPUTER .........& KL Ahxiad. M . Colenso. and G . Corbett THE VALUE OF THE COMPUTER IN EDITING .................................... AN 'OPEN TRADITION' TEXT J.R.C. Martyn PROBABILITIES OF OCCURRENCE OF CHARACTERS. CHARACTER.PAIRS. AND CHARACTER-TRIPLETS . IN BNaISH TEXT ............................ R Shinghal and G.T' Toussaint SENTENCE LENGTH DISTRIBUTIONS IN GREEK HEXAMETERS Ahl HOMER ...... S . Michaslson. A.Q. Morton. and W.C. Walce SNOBOL: THE LANGUAGE FOR L'ITERARY COMPUTING .................... L.D. Burnard GENERATING AND TRANSFORMING BY A COMPUTER </SectionTitle> <Paragraph position="0"> Appendix XI. Properties of Lex+cal Rezations.</Paragraph> <Paragraph position="1"> a, Refldvity, Symmetry, Transitivity.</Paragraph> <Paragraph position="2"> Certain properties of lexical-sewtic relations can be very useful in deductive inference. For instance, 15 we know that a cheetah is a ki,na or mammal anu a mamm i.s a kind of vertebrate then we can deduce that a cheetah is a kind of vertebrate. Writing T for the taxonomy relation, we can abbreviate this sentence: if cheetah T mammal and mammal T vertebrate then cheetah T vertebrate. Whenever bTc and cTd, it follcaws that bTd. This fact ran be described much more effi ciently by the stuement that the taxonomy relation is transitive. Two other commonly menttand properties of relations are refilexivity and syrmnetry. These properties may ppply to predicates formed from lexical entries as well as to lexical-semantic relations.</Paragraph> <Paragraph position="3"> To be precise, a relation R defined on a set S is said to be a trana~t<ve relation if whenever b and c are R-related and also c and d are I? related then b and d staAd in a relation R also, Synonyniy is a transitive relation just as transitivity is. The preposition in behaves in the same way. If Sam is in the kitchen and the kitchen is in the hotel, then we know that Sam is in the hotel. The time interrelation before behaves like this, too. If Zorro arrived before the posse did and the posse arrived before thz explosion, then we know thgt Zorro arrived before the explosion.</Paragraph> <Paragraph position="4"> A relation R defined on a set S is said to have the refZez<ue property if all the elements of S are R-related to thenl~elves, that is, if mRm is true for all members m of the set S, The synonymy relation has this property a word means the same thin% as itself. The antonymy relation ANTI does not have this property. It is not rrue tha&, hot ANTI be, for example.</Paragraph> <Paragraph position="5"> A relation R defined on a set S is said to be e~stric if whenever,b and c are R-related then so are c and b; that is, R is symme.tric if and only if bRc always implies cRb. Synonymy also has this property. If b is synonymous with c, then c is synonymous with b. So has antonfly. Given that hot ANTI sold, we immediately know that= cdd ANTI hot. Taxonomy ie not eymmetric, however. A lion is a kind of mammal, but a mammal is not a kind of lion.</Paragraph> <Paragraph position="6"> In question answering we may be just as Interested in drawing negative conclusions as positive-ones. Thus i~rmay be important to know tliat tf bRc is true then cRb must be falae. The term asynmrstrio is used to describe a relation R for which bRc and cRb are never both true, at $east when b and c are different elements of the stt S. Taxonomy is asymmetric and so is the thug interrelation before. If the question asks, &quot;Did c happen before b?&quot; and we know that b happened before r, we can answer with a confident no. For want of a better term we will say that the relr Sion R is mn-synonetrio if it is neither symmetric or &symmetric. In this case bRc and cRb are sometimes both true and sometfmes not. Shilarly, he term imefz.exive is used for the case in which mRm is never true, while the term nonreflexit)e is used for the case in which mRm is sometfmes true and sometimes not. In the same way intransiti~e is taken to mean that if bRc and cRd, we can conclude that b and d are not R-related, while nantrcrnsitive will mean that bRd is sometimes true if bRc and cRd, but- not always.</Paragraph> <Paragraph position="7"> Each lexical relation itself; has a lexical entry. The reflexivity, symmetry, and transitivity properties of the relation are listed in this entry, as they are in the entries for interrelational operators and prepositions and other lexical item for which they are relevant.</Paragraph> <Paragraph position="8"> There are also lexical entries under the property names, refldvs, irr~~~~vr, etc. listing the appropriate axioms. The motivation behind laical entries for properties is first of all greater generality.</Paragraph> <Paragraph position="9"> Secondly, it makes it much easier to add lexical relations and to add other properties which turn out to be useful.</Paragraph> <Paragraph position="10"> At this stage of development there are several transitivity axioms: For lexical relations Rel, like taxonomy b Re1 c c Re4 d ' b Re1 d For interrelations J, like bsfbre lloZdsfI(~,~,,Z,)) A RoZda (I(J, Z2,Z3)) Holds (I(J,Z~ ,z$) For prepositions Q like in or abave Intuitively these are all instances of the same concept, transitivity. Theze should be some single way of expressing it. It is a defect of this representation system that there is not.</Paragraph> <Paragraph position="11"> A relation that is reflexive, symmetric, and transitive Ls called an equivalence relation. The synonymy relation is an equivalence relation since it has all three properties. If R is an equivalence relation, then a subset consistrlng of all the elements which are R-related to a partirtllar element x by the7equivalence relation is called an equivalence class. In an equivalence class all the elements are R-related to each other. An equi vaf ence relation partitions a set into equivalence classes ; each element of the set belongs to exactly one equivalence claas. The synonymy relation paftitions the items in the lexicon Ln just this way. There Is a class consisting of stcsp%oion and all the words synonymous with 8148phion, like mistmcet and dozibt. These synonymy classes are disjoint; each word sense in the lexicon belongs to exactly one of them (cf.</Paragraph> <Paragraph position="12"> Edrrmndson and Epstein 1972, Palmer 1976).</Paragraph> <Paragraph position="13"> With this ge a basis an equivalence relation of paxaphrasability between sentences can be established. Sentence S1 is a paraphrase of sentence S2 if one is obtained from the other 3y substituting synonyms for each other.</Paragraph> <Paragraph position="14"> d Mr. Kennedy viewed Lady Laura with suspicion.</Paragraph> <Paragraph position="15"> Mr. Kennedy regarded Lady Laura with mistrust.</Paragraph> <Paragraph position="16"> We might also allow substitution of conversives, nominalizations, etc. Nancy was Sally's student.</Paragraph> <Paragraph position="17"> Sally was ~ancy's teacher.</Paragraph> <Paragraph position="18"> Sally taught Nancy.</Paragraph> <Paragraph position="19"> The equivalence classes of this relation, each one of which is the set of all pamaphrases of a given sentence have a definite theoretical importance and some practical significance in question answering. One member of a class might well 'be part of the story; another the right answer to* a question.</Paragraph> <Paragraph position="21"> This representation system can be viewed as defining a relation P such that S1 P S2 if and only if S1 and S2 have the same representation. null If the representation system is well defined, then P should define the same equivalence classes as the paraphrasabilitv relatior b. Xttt)srses, The inverse R of the relation R is the relation which &quot;goes in the opposite direction&quot; from R; that is, bRc if and only if cRb. Thus, bake T make and mke T bake are two ways of saying the same thing. Both pieces of information are stated in the lexicon.</Paragraph> <Paragraph position="22"> However, the lexical entry for Eake includes T nuke; the lexical entry for naks includes T hake. Why bother to say the same thing in different places? There are two reasons for this. First of all, the inversa relation may be a relation that is conm~nly and easily verbalized, worth naming in its own right. This is certainly true of the CHILD relation, as in pu~py CHILD 20g. Instead of asking &quot;What is a baby dog called?&quot;, we could ask ''What is a grow? puppy talled?&quot; or &quot;What does a puppy grow up to be?&quot; The second reason is that putting this information in both entries can &e searches easier and much faster, We may only have one half of the pair and need the other. We may have dog and pppy. This is easy if we have the information CHILD pppz~ in the dog entry. Othewise we might have to search the whole lexicon, In other situations we have two words but no direct connection between them. For example, suppose the system knows twn T mama2 and maZ T vertebrate and is then asked, &quot;Is a lion a vertebrate?&quot; The connection betwen Zion and vertebrate can be found much more quickly if the search starts *om both the vertebrate end and the Zion end of the chain at the same time, but to do this there must be s pointer to m~mmaZ in the oertebrute entry.</Paragraph> <Paragraph position="23"> Another question comes -to mind. Why call the inverse relation to CHILD by the clumsy name CAaD instead of its propel name PARENT? The ECD uses t~o different names for a relation and its inverse (So and Vo ace inverses, for example). If this were dane here, two versions of the appropriate axiom schemes would be needed, one in the CHILD entry and one in the PMNT entry. Since a relation R is called symmetric if bRc alwaye implies cRb, it follows that a symmetric relation ie its own inverse.</Paragraph> <Paragraph position="24"> The syaonymy relation S and antonymy relation ANTI are both self-inverse in thie sense.</Paragraph> <Paragraph position="26"> For this reason we never need the spnbol ANTI, etc. ANTI is MITI The entry for hot includes ANTI cold, the entry for cold includes ANTI hot.</Paragraph> <Paragraph position="27"> 0. (hrique Linkage.</Paragraph> <Paragraph position="28"> Raphael (1968) has proposed a property which seems extremely useful. He calls it m6qus-Z$nkuge (U). Nathematicians usually.refer to such relatdons as one-to-one. A relation R has the unique-linkage property if whenever xRy then bRy is false for any bk and xRc is false for any cry, i.e. any object is R-related to at most one other. ~aphael's example of unique-linkage is the relation &quot;just to the right of&quot;. The behavior ie especially characteristic of the queuing relation, e.g. with days of the week, Monday Q Tuesday, etc.Some relations may be uniquely linked on one side only, e.g. mother-child is uniquely linked on the left. We can define UL unique-linkage on the left and UR unique linkage on the fight. (A relation which is UR is a single-valued function. If R has the UL property, then its inverse is a single-valued function.) Raphael also proposed for SIR-1 (ibid, p. 101) a property which he calls ixreflexive. R is set-nunreflexive if (\lx M)--WBcX) 6!acX) @RBI In the SIR model both the 'X is a part of Y' and the 'X is owned by Yf relations hwe this property. What $t qays is that every set in the model has a minimal element with respect to the relation R. A siapler version of th&s property is sufficient for our purposes.</Paragraph> <Paragraph position="29"> Minimum ~cM) - ()'Y i X) (32 X) {ZRY) Condition Every noneslpty subset has a minLmum.</Paragraph> <Paragraph position="30"> Maximum WXcM) -- (qr X) (3 Z X) (YRZ) Condition Every nonempty subset has a maximum.</Paragraph> <Paragraph position="31"> The part-whole relation' has both properties in our model. In any non-empty subset in the model there is something in it that is not a proper subpart of anything else in that subset, and also something that has no proper subpart. A relation that has this property stops samewhere. It is not reflexive and not circular, A search that goes on looking for links of this kind will stop somewhere. The relation 'is an ancestor of' has this property. We will eventually run out of ancestors in one direction and descendants in the other, at least, inside a finite model. The properties of relations are summarized in Table 4.</Paragraph> <Paragraph position="32"> on the left WEM) WY M) (XRY -' WZ EM) (ZRY ' Z=X) ) uniquely linked on the right wXEM)(4/Y fM)XRY4 WZ EM){XRZ4 Z=Y)) d. Pmtial Ordering.</Paragraph> <Paragraph position="33"> Any transitive relation defines a partial ordering. Several of the lexisgl relatiohs discueeed earl,ier are transitive; many lexical items are transitive too. One important reason for.repr*senting time in terns of the transitive interrelatEton before is to allow one to make the same kinds of sdmple deductions about time that one can make about taxonomy. Some transitive relations, like taxonomy, are alsb reflexive. In this case we talk about a weak order<ng. (X s: Y for numbers is a weak ordering.) Some are not reflexive, these are called strong ordering relations. (X < Y for numbers is a strong ordering.) The time relation before 5s a strong order3ng relation. For any weak ordering there is a strong ordering and conversely, Starting with the taxonomy relation T,</Paragraph> </Section> <Section position="8" start_page="1" end_page="1" type="metho"> <SectionTitle> 1 I </SectionTitle> <Paragraph position="0"> for example, a relation TI or proper,taxonomyl' can be defined consisting of the pairs x and y for which xTy but x and y are different. Then~Tly means that x is a kind of y but different from y. If instead one starts with strong ordering relation before, one can deane a weak relation &quot;beforel&quot; for which x beforel y means that either x before y or x cooccured with y.</Paragraph> <Paragraph position="1"> The queuing relation Q is nat itself a partla3 ordering but a partial ordering can 'be derived from it. Monday Q Tuesday anLTuesday Q Wednesday, but it is false that Monday Q Wednesday. Queuing is an 'immediate successor *lation like the relatxon between a natural number n and the next number n+l. A relation Q' can be defined such that xQ1y if either xQy or there are some objects cl,z2,. ..,z, such that xQzl, z1Qz2, ... znQy. It follows immediately that if bQc and cQd then bQ1d. Q', the 'successor' relation, 95 2 4 ie now transitive, for if #ltc and cQ1d, then one can find s chain of Q-related objects linking b and d just bv cbncatenating the chain linking c and d, Rapbl's pair of relations jright and right behave this way. The relations &quot;is a child of&quot; and &quot;is a descendant of&quot; are alga pafred in this way,</Paragraph> </Section> <Section position="9" start_page="1" end_page="6" type="metho"> <SectionTitle> SUMM (9 l$X </SectionTitle> <Paragraph position="0"> Ordinary dictionaries have not been given their due, either AH 80ur~e8 of material for natural language understanding syetema or as corpora that can be used to unravel the complexities of meaning and how it is represented. If either of these goal8 are aver to be ~chieved, I believe that investigators must develop methods for extracting the semantic content of dictionaries (or at least for transforming it into a more useful form).</Paragraph> <Paragraph position="1"> It is argued that definitions contain a great deal of information about the semantic characteristics which should be attached to a lexeme, To extract or surfacke such infarmation, it will be necessary to systematize definitions and what they represent, probably using semantic primitives. In this paper, I deecribe procedures which I have developed in an attempt to accomplish these objectives for the set of verbs in Websterle Third New Intern~tional Dictionary (~3). I describe (4) how I have'used the structure of the dictionary itself in an attempt find semantic primi tive s and how appears that the systematization must incorporate a capability for word sense diecrimination and must capture the knowledge contained in a definition.</Paragraph> <Paragraph position="2"> The body of the paper is concerned with demonstrating that semantic information can be surf aced through a rigorous analysis of dictionary definitions. The first step in this process reavires- a clom~phenaive framw~ark- within WkLch def iait ions can be an~ly~ed. In,dcvelopinp thls framework, we must r~membrr thqt ~qch wordlu~erl. in I definl tion is .ilm dc1 ineci in the rl~ctionqry, so that we must be qble to uncsvpr ~nd dc..~? kit!! v1cious circles, The framework must llso be cwable oi rerrt-senting traditionql nations of q~nerative grammar to de3l wiTh the syntnct~c structure of definltlon~, s~ritable framework ,IF-Pears to be arovided bv the theory of lqbrled directrd (T~PILP ( di,graphs) .</Paragraph> <Paragraph position="3"> Using points to represent dictionqry entries ~nd lines to represent the relation &quot;is used to defi.neV, two models of the dictionary are described. ?rro theee models and from digrwh theory, we cqn conclude that there may exist orimi-tive units of meaning from which 911 concepts in the dictionary can be derlved.</Paragraph> <Paragraph position="4"> To determine arimitive concepts, it is necessarv to sub-ject definitiuns to syntactic and semantic nsrsinp in order to identify characteristics that should be att~chkd to each definition. Syntactic parsing such as that implemented for systemic grammar by Minograd is the first stea. semantic parser must next be developed. Tt appears that definitions themselves, and particularly definitions of prep~si.ti~ns (which are used to express sense relations), will be of sipi,ficant help :in develop; ing such a Darser, Further work is necessary to develon procedures for surfacing from definitions i.nformation about the context which must be associ.ated with each sense. It wpears as ib this Darser wlll have more ~eneral use for ordlnary discourse.</Paragraph> <Paragraph position="5"> These notions lead to the ultimate model of a dictionary, where points represent concepts (which nay be verbalized and symbolized in more than one lay) and lines represent relations (synta~ti,c or aemantk-c) between canoepts.</Paragraph> <Paragraph position="6"> Ba ~ed on these models, procedures for f i,nding prirniti-ve concepts are described, using the set of verbs and their definitions from W3. Specific rules are described, based on some elementary graph-th6qre tic principles, structural characteristics of dictionary de'finitiohs, and the parsing of the definitions. These rules have thus far reduced the initial eet of 20,000 verbs to fewer than 4,000, with further reduction to cone as all rules are applied, It is argued that this approach bears a~ strong relationship to efforts to represent knowledge in framecr. Although much work is needed on the parser and on a computerized version of this approach, there is some hope that the parser, if expectations are borne out, will be capable of transforming ordinary discourse into canonical frame representations,</Paragraph> </Section> <Section position="10" start_page="6" end_page="10" type="metho"> <SectionTitle> 1 . INTRODUCTION </SectionTitle> <Paragraph position="0"> During the pa~t 15 years. scientists in many fields have been building a reservoir of knowledge about the semantic char acteristics of natural language. Perhaps somewhat inexplicably znese developments have for the most part Agnored the semantic contenl of dictionaries, despite the fact that even a small one contain8 a vast amount of material. Some attempts have been made to dent these repositories, but the steps t'aken have been tentative and have not yet borne significant fruit, perhaps because che sheer volume and scope of a dictionary is so overwhelming. As a result, most studies have dealt with only a few definitions wj%hout a comprehensive assault on the whole. While such studies have led to many insights, it seems that the full ugerulness of a dictionary's cantents will be realized only when a comprehensive model of its semantic structure is dweloped, null Any system intended to provide natural language understanding must necessarily include a dictiona~. If any such system is to achieve broad applicability, its dictionary lnust cover a substantial pat of the natural language lexicon. For this to occur, the developers of a system must either create a dictionary from scratch or be able to incorporate an existing dictxonary. Given the amount of effort that usually goes into development of an ordinary dictionary, the former a1 ternative is rather impractical. Bowever, little has been done toward meetinn the latter alternative; with wnat follows, I will describe the approach which I believe must be followed in transforming the contents of an ordinary dictionary for us6 In a true naturaX language system, In order to be used in a language understanding system, a dictionary's semantic contents must be systematized in a way that the sense in which a word ia being used can be identified. Bbfore thi~ can be done, it is necessary to characterize what 1s already cantained in each definition. To do this, it seems necessary to write the meaning of each definition in terms of serpantic and syntac5ic primitives. My purpose in this paper ia (1) to describe how to use the dictionary itself to move toward idhntification of the primitives, at the same time (2) showing how this process can be used (a) to provide the capability for discriminating among word senses ( i. e. characterizing; the frames into which a given word sense will fit) and (b) to characteriee knowledge contained or presupposed in a definition. Before elhbarking on the description, it-$8 necessary tc paint out some limitations whZch shaad be kept in mind as Dhe reader proceeds. First, in trying to @resent an overview of my approach, I have had to forgo describing the detailed steps which I have followed to date. Second, even had I presented a full description, I would still have been short of providing sufficient details to enable computer implementation of any procedures. Third, Since the approach presumgs that cancepts represented by the lexicon are tne realizations of many as yet unknovin-rrecursive functions to be dl scovered by stripping away one-layer at a t~me, results other than procedures to be used An stripplpg will not emerge untll all layers have been removed. (However, I do wrae that the llstripplngm procedures are inherently useful, in that they will constitute a parser even in the intermea~ase stages.) Fourth, since I have not ha@ access to a computer, which has become essentlalLfor significant further progress, I have been unable to determine how far the grocedures I have developed would take me, so there iLs an inherent uncertain-ty as to how much further development as needed. Notw~thstandlng these limitations, I am hopeful that what is prenented will provide a satisfactof.y framework for further iLnvestigations into the contents of dictlonarles. I will comment further on these limitahions and how they might be overcome at the end of the paper.</Paragraph> <Paragraph position="1"> 2, ATTITUDB'3 TOWARD DICTION4RIE5 Many of +he siqnifxcant contributors to the present understandxng of rneanlng (such as Xatz and Fodor 1963, Plllmore 1968 and - 7971, CHafe 1970, Jackendoff 1974, wlnograd 1972, and Schank 9972) have generally lgnored dictlonarles. Yet, each has presented a formulai~ structure for lexical entrhe5 to serve as a bas= for the creation of a rlew dictionary 4lthough their perceptions abouti the nature of language are well-established, thei? formellsms for lexxcal entries have not taken advantage of the equally well-establ~shed praetlces of lexicography.</Paragraph> <Paragraph position="2"> The rationale underlslng the development of new fommalisms~ ex~rer~sed in some cases and ~m~llcrt t;n others, ids that lexlcal entries in dictionaries am unsatisfactory DeCAuse they do not contain sufficient infomation. These formali-sms thus require that semantic features such as 1lanirnateft or &quot;statew be appended* to particular ent*ies. While it is true that ordinary diotionary entries do not overtly identify all appropriate features, this may be lees a dlfficulhg inherent in definitions than the fact thst no one has developed the necessary mechanisms for surfacing features from definitions. Thus, for examp3.e. ltnurse1' may not have the feature llanLmatew in its definition, but t?nuraew is defined as a ltwomanw which fs defined ad a tlpersonw ~hich is defined as a 1&quot;beingfl&quot; which &quot;Ys defined as a &quot;living thingw; this string seems sufficieht te estabaish &quot;nurseN as &quot;anirnatell. In general, it seems that, if a semantic feature is essential to the meaning ofa particular entry, it is similarly necessary %Hat the feature be discoverable within the semantic structure of a dictionary, Otherwise, there is a defect in one or more definitione, or the dictionary- contains some internal inconsistency. (Clearly, it is beyond expectation that any pre~nt dictionary will be free of these problems.) The possibility of defective definitions has also gene^-ated crf ticiams, more direct than above, on the potential usefulness Of a dictionary. On one Hand definitions are viewed as &quot;deficient in the presentation of relevant dataw since they provide meanin- bv ueing &quot;substitutable words (i.e. by synonyms), rather than by listing distinctive femtureafl (Nida 1975 : 172) . On another hand-, the proliferation of meanings attached to an entry is viewed as only a case of &quot;apparent polyeenyN which obscures the more general meaning of a lexeme by the addition of &quot;redundant features already determined by the environmentft (Bennett 1975:4-1.1). Both objections may have much validity and ts that extent would necessitate revisions to iqdividu& or sets of definitions. However, neither viewpoint is sufficient' to preclude an analysis of what actually appears in any dictionary. It is possible that a cbmprehensive analysis might more readily surface such difficulties and make their amelioration (and the consequent improvement of definitions) that mu& easier, Xven though dictionaries are viewed somewhat askance by many who study meaning, it seems that this viewpdint is influenced more by the difficulty o* systematically tapping their contents than by my substantive objections which conclusively establish themas ~seless repositories of semantic content.</Paragraph> <Paragraph position="3"> However, it is necessary to demonstrate that a spstematic app~oach exists and can yield useful results.</Paragraph> </Section> <Section position="11" start_page="10" end_page="12" type="metho"> <SectionTitle> 3, PREVIOUS RESXARCN ON DICTXONARIES </SectionTitle> <Paragraph position="0"> Notwithstanaing the foregoing direct and indirect criticisms. some attempts have been made to probc the nature and& structure of dictionary definitions. A review of relevant aspects QI- two such studus will help the niaterial presented here stand out in sharper relief.</Paragraph> <Paragraph position="1"> Olney 1968 describes the conceptual baais of many pro$eetted routines for processing a machine-readable transcript of Webster ' s Seveqth New Collq&ate Dictionary (~7). The primary objectives of these routines were the development of &quot;(a) rules for obtaining c-ertain of the senses described for W7 entries from other senses described for the same entries or from senses described for other W7 entries from which the first (at least in typical cases) were derived morphologically; and (b) semantic wmponents and rules for combining them to yield specifications of senses that cannot conveniently be obtained br rules refer~ed to in (a) above.&quot; (ibid. : 6) Although these objectives me reasonable, they do not take advantage of the possibility that the semantic structure of a dlictionary might be a unlfied whole. As a\ result, an8 routines that are developed seem to require the serendipitous perception of patterns. Further, i0 a dictionary does have a unified semantle stpucture, it is not clear that a rule relating meaning to form wil-1 be relevant toga model' of the semantic structure even though interesting results might emerge. It seems n-cessary to have some comprehensive view that will permit un to kaW whether a particular rule is well-formed. This lack of objective criteria also im~erils any anaIysis- that selects a sub-set of definrrions for detailed analysis. The selection of a subset of the dictionary shoulcl. arise from wll-defined a priori considerations mmer than an intuition that a particular wbset seems to be related, An example of this intuitive agproach appears ~JI Simmons 1975 and 1976.</Paragraph> <Paragraph position="2"> rn Quillian 1968, the analysis of dictionary definitions was part of a study of semantic memory, and for that reason was noP concerned with the full development of a dictionary model. In that study, a person determined the mesning of a concept when he &quot;looked up the 'patriarch1 ward in a dictionary, then looked up every word in each of its definitions, then looked up every word f6hnd-in each of those, and so on, continually branching outward until every word he could reach by this process had been Looked up once.&quot; This process was never actually carried out because (1) not all words in a dictionary were used in the computer files, (2) the process was terminated when a common word was found in comparing the meanings of two words, and (3) there was a bellef that there are no primitive ward concepts. The termination of a search 3x designed was necessary in any event since, without my restrictions, it is likely that a large part of the dhztionary would have been reached on every occasion, More importantly, Quillian did not fully consider wHat was happening when branching led to a word already encountered, namely, that a definitional circularity was thereby uncovere6 Such circularities which mi-ght be vicious cir-cles, must be treatea specially (as will be shown below), and hence, Quillian8 s unrestricted branching should have been mdifbed.</Paragraph> <Paragraph position="3"> Quill ian also overlooked the. possibility that a concept common to two qatriarchs is more primitive than either. The continued comparison of more and more primitive concepts, along with restricti~ns on the outward branching, implies that primitiive concepts actually do Based on these observations, I take, as a working hypothesis, the assumption that a dictionary may be a unified whole with underlying primitive conce~ts. ' With thin beginning, it is necessary to articulate a mod& of the dictionary which will permit an identifiqatian of the primitive concepts through the application of well-defiaea rules or procedures. It is proposed that what follows constitutes the first steps toward meeting this objective,</Paragraph> </Section> <Section position="12" start_page="12" end_page="16" type="metho"> <SectionTitle> 4. DXSRIPTIBN OF -- UICTIONRHY . . CONTENTS </SectionTitle> <Paragraph position="0"> Since a dictionary contains much material, it is first necessary to delineate exactly what is to be modeled-? For thi~s purpose, it is assumed that the semantic content of a dictionary essentially resides within its defi.nitions, thereby excluding from formal analysis such things as the pronunciation, the etymol~gy, and illwtrative examples. s presently concelvea, the analysis will focus on the ward belng defined (hereafter called Ehe main entry) , the definitions ( including sense numbers and letters used as delimiters) , part-of-speech I No dictianary is likely ta satisfy thls assumption, which is only a theoretically desirable characteristic. The assumption enables us to exclude the definienda from the models, In the interests of space, I have glossed over B large number of intricacies that would have to be dealt with in arriving at a machine-readable hnscript suitable for analysis.</Paragraph> <Paragraph position="1"> Several pages would be reqyired to describe them fully.</Paragraph> <Paragraph position="2"> labels, status or usage labels, and usage notes. The manner in which these features will be employed will be made clear as the analysis proceeds.</Paragraph> <Paragraph position="3"> The hypothesized unified nature of a dictionary arises from the fact that definitions are expressed by werds which are 4180 defined (i., there is no semantic metalangua~e). If we wish to understand the meaning of a given definition, then we must first understand the meanings of its constituent w&dse Since each constituent corresponds to a main entry, then, in order to understand the meaning of the given definition, we mus% understand the meaning of the constituent wards1 definitions, Continued repetition of the process is nothing more than , the outward branching process described by Quillian; however, as mentioned before, we must make this branching more disciplined in order to deal with vicious circles and avoid unwanted circularities, If we are to have a fully consistent dictiona~y, its model must show how each definition is related to all others. Thus, for each definition, X, the model should enable to identify ( 1) those definitions of the constituent wordr of X that apply and those that do not apply, and (2) the production rules that generated X from these definitions. For exampl,e, in the definition of tqe noun broadcast, &quot;the act of spreading abroadu, 4 it There are some exceptions to this assertion, such as groper names, . biological caxa, and other special symbols, a s pointed out by the Journal's referee.</Paragraph> <Paragraph position="4"> is necessary that the model indicate (1) which of the definitions of --- the, act, of, spread, and abroad apply, and (2) the production rules by which - the and ___I_ act and all other collocations) occur together. If this can be done for each definition in the dict~onary, and if any inconsistencies are reconciled, then, as will be shown, it should be possible to find the primitive concepts in the dictionary and to transform each definition mto a canonical f Drm.</Paragraph> <Paragraph position="5"> 5, - BJSIC MODEL The theory of (labeled) directed graphs (digraphs)5 is used as the formalism for the modds. Digraph 'theory deals wj th the abstract notions of lfpointsff and &quot;directed linest1 ; its applicability to the problem before us therefore depends on how these notions are interpreted. In this respect, it 1s important to distinguish tpe manner in which this theory is used here from the manner in which it previously has been used in semantics and linguistics. The two most common uses are (1) where trees display phrase and syntactic structures (cf. Kate and Fodor 1963), or (2) where directed graphs portray the seguena tial generation of words in a se~tence or phrase lcf. Simmons 1972). In these cases and others (cf. Quillian 1968 and Bennett 1975) graphs are used primarily as a vehicle for display All definitmns ueed in this paver are taken from Websterts . Third New International ~iction&ry, Eficyclopaed~a Britannica, Chicago, 1965.</Paragraph> <Paragraph position="6"> Terminolqy for digraphs follows Rarary 1965.</Paragraph> <Paragraph position="7"> and no results from graph theorv are expPicitly employed to d>aw further inferences. However, as used here, g~aphs consti tute an essential basis for the analysis and hence will play an integral role in a nulrrber of assertions that are made.</Paragraph> <Paragraph position="8"> In the simplest model, a point can be interpreted as representing all the definitions appearinpunder a single maln entry; the main entry word can be construed as the label for that point. The part-of-speech labels, status or usage labels, and usage notes are considered integral to the definitions and may be viewed as part of a set of characteristics of the individual defxnitions. A directed line from x to y will be used to represent the asymmetric relation &quot;x is used to define yu; thus, if the main entry x appears exactly or in an inflected form In a definition of y, then xRy. (This does not preclude a distinct line for yRx or XRX.) Therefore, we can establish a point for every main entry in a dictionary and draw he appropriate directed lines to form a digraph consisting of the entire dictim nary. (~hls digraph may be disconnected, but probably is not.) An example., which 1s a subggaph of the dictibnary digraph, 1s shown in Figure 1 on the next page. Xxcept for broadcast, only the labels of each point are shown, but each represents all the definitions appearing at its respective main entry. The directed line from - act to broadcast corresponds to the fact tha* &quot;act is used to define broa@castn, since its token appears in &quot;tfle act of spreading abroad&quot;. In this model, the token &quot;spreadingH is not represented by a point, since it is not a main ertry.</Paragraph> <Paragraph position="9"> broadcast (the act of spreading abroad) digraph using the baszc model.</Paragraph> <Paragraph position="10"> Since the definition shown iLs not the only one for broadcast, thls point has additional ancorning lines which ape not shown. The resultant digraph for even a small dictionary i.S extremely large, perhaps consibsting of well over 100,000 points and 1,000,000 lines. Clearly, such a digraph provides little fin& structure, but even so, it does have some utility. The manner, i.n whdch it can be used is descr,i.bed 9n Section 9.</Paragraph> </Section> <Section position="13" start_page="16" end_page="38" type="metho"> <SectionTitle> 6. EXPANSION OF THL MODXL: POIN_S 45 DEFINITIONS </SectionTitle> <Paragraph position="0"> Lett5ng each poi.nt in the basic model represent all the definitions of a main entry provides very lfttle del?neatAon of subtle gradations of semantic content. As a first step toward understanding this content, it seems worthwhile to let each point represent only one definition. However, the basic model will not trivially accommod&te such a spec~ficataon i~rimarily because of the interpretation gzven tg the directed llne), and thus it must first be modified, In the basic model, the exzstence of a line between two points, x and y, assertr that xRy, I., &quot;r 1s used to define yB1. Sfnce the points represent all the Cieflnltlons under the main entries, the existence of a line arises from the simple fact that x appears in at least one of yes definitions. ff the point y represents only one definition, say y , there 4s no dlfflculty in saylng that xRyj. However, if we wf sh every polnt to represent only one definltlon, then we must frind the deflnltlon of x, say xl, for whlch xlHy is true. Referrinp to the subgraph An Figure 1, this amounts to determining, for example, which def-inition of abroad b is used to defi.ne the token ltabroad&quot; inn &quot;the act of spreading abroadN, that js, finding the i such that &quot;abroad,Rthe act of spreading ab-roadfl or It should be intuitively clear 'that 3nterpretation of points as amgle @efPn.itfons is desirable. However, there are no a prior1 crateria by which the appropritate value of i can be determhed, and hence there is no immediate transformation of the basic model hnto a model where each mint represents one qefinition. Sance th~s objective is wollth pursuing, it 3s therefore necessary to develop criteria or rules according to which the desired transformation can be made.</Paragraph> <Paragraph position="1"> In the appli.catAon of rules that may be deueloped, it will be convenient to make use of a model intermediate between the basic one and the one atlth points as definitions. For this purpose, we can comblne the two models by employlng a trivial ze lation, xLRx, which says that the ith defAnltion of x is used to define x; this holds for all definitions of x. The line reflecting xRy would remain in the mqdel, so that the digraph (the act of spreading abroad) both eingle and multiple defiqJ tiom.</Paragraph> <Paragraph position="2"> would show both xlRqdand xRy and x would l5e a carrier, .as ilj null lustrated in Tigure 2. In this case, the unsubscripted abroad represents all the definitions of I abroad (olny some of whia are shown). If and when suitable criteria establish, for example, ikat abroad,, but not abroad --2 * abroad ,..., fits chc context of the token llabroadn in-thr definition of broadcas i, it would then be possible to draw a line directly from ?broad1 to broadcast without the intermediation of the ansubscripted Roint abroad, thus eliminating* paths from abroad*, abr~ab~, . . ,* This model thus includes the points of' the basic model and adds mints to represent each individual definition in the dictionary. The lines betwen these points ensure that no relatian in the basic made1 is lo&*. As described in the example, 'it is necessary to develop rules according to which the points representing more than one definition can be eliminated or bypassed, 80 that the Only relatl ons, xRy, that remain are such that x and y are poi~ts which represent one definition, It way happen during the application of rules that some lines to a carrler will be eliminatgd with more than one st111 remaining. In such a case, it will still be useful to modify the digraph as much as possible. For example, if xRy in the basic nlpdel, whepe x has m definitions and y has n, and xRy is the ejrpanbd model, then x, ,. . . ,xmRyJ. It may be that mme Crkterion indicates that, say x, ,x2Ryj but that xg . , x,Ry When j' this occurs, we can create points xa and xbesuch that X, ,9c2flxa xaRyj, and x . . ,x Rxb, but with no line from xb to grouping will be demonstrated in Section 9. In any event, since maw criteria will eventually be requj red in the elimination of points representing two or more bfinitions. this abklity to group definitjons is a necessary mechanism for modeling intermediate descriptions of the dict~onarj. (It should be noted here that all such points will not be elimina fed; those that remain will indicate an essential ambiguity in the dictionary; this is further discu%sed in Section 8.) 7. SEMANTIC, STRUCTURAL, AND SYNTACTIC PARSING OP DEFINITIONS The basic ad expanded models, exampled in Figures 1, 2, and 3, do not portray any of the meaning of the di~tiohary, but rather indioate where particular relationships exist. In fact, these two models portray only the relation &quot;is used to definett as if there is no other relation between definitions. This approach does not capture some very important elements that go to make up a definition.</Paragraph> <Paragraph position="3"> Instead of being analyzed directly into its ultimate constituents, a6 in Figures 1 and 2, the definition, &quot;the act of spreading abroad&quot;, should fir-.,st be br*en down into sub~hrases and then into its ultimate qnstf tuents, s s in Figure 4, shown on the next page. A-desirable property of the new pointe is that they have the syntactical structure ox derinitions; Thus, ff the act&quot; and ftspreading abmad&quot; have the form of noun def initionsy &quot;spread abroad&quot; has the form of a verb definition: and nof spreading abroadN (not shown, but feasible under a diTfesent parsing) has the form of an adjective definition. This would elfminam such combinations as &quot;act afll or the&quot;. The poinss represen-ung pbase consti-tuenta of a def-bni ti on thus have the form of definitions, but lack a label.</Paragraph> <Paragraph position="4"> The absence or presence of a label seems to make no difference in understagding the definition represented. In fact, (the act of spreading abroad) it seems val.id to represent identically worded definitions or phrase constltuenfq, regardless ~f the number of main entries under which they appear, by a single point with multiple labels. Thus, if each of the main entries disperse, scatter, and A diatribute has a definition verbalized as spread abroadtt, these three words can be labels of the point lfspread abroad&quot; jn Figure 4. auch a construction has no effect on the analysis of 'the definition &quot;the act of spreading abroad&quot; or &quot;spread abroad&quot; as showr in F~gure 4, and si,miiarly, the analysis there would have no effect on any analysis involv ng disperse, scatter, or - distribute. Since thtre is a large number of fnstances where duplicate wording appears in a dictionary, the approach given here would effect a substantial reduction in the she of the digraph. (This is not to say that the words diapePse, scatter, and distribute haua ae same meaning, but rather that in some instances these words can express the same concept.) The definition, X, &quot;the act of spreading ,ibroad!' is essentially an entity unto itself. The definitimns of its component words have similar independence, However, lkke atoms in molecslles, we need to identify those forces which hold the components %ogether a~ld which endow the whole with whatever chara'ct3ristlcs it has. The d@ finitions of the component words may require several worde for their expression. but'thev are symeoliwd %y one word in the definition X; even so the symbol and the definition both represent the same entity, which has certain charactefisCics enabling it to be acted upon by ce~tain forces. These characteristics are the semantic, ~tructural, md syntactic properties of defihitions, and the forces are the production rules by which the entities (i. the component definitions or their symbols) are brought together. A definition may be viewed as the realization of such rules operating on the chgiraeteristics of other definitions. The nerculean task before us is to build a parsing system or recognition grammar which ill articulate the e%xracteristics %o be attached to each def-inition and which wul capture the production rules necessary to portray the relationships between definitions. The remainder of this section will present my ideas on how to approach this task.</Paragraph> <Paragraph position="5"> The pT.ocess which I have used IUL. finding primitives entails showing that one definition is derived from another thereby excluding the former as a candidate for being primitive. Such a demonstration of a derivational relationship requires a parser. Each pattern which I observe b6 tween definitions helps to exclude fu~ ther definl tions and simultaneausly becomes part o? the parser. As 2 result, identincation of the charatteristics Lo be attached to eacfi def~nition does not have to *be accomplished all at once; as will bacome clear below, our purposes can be served as the components of thc parser sre de-</Paragraph> <Paragraph position="7"> Ilneated. Thug, success does not require trill n~ticulat~nn of the parser before any parsing is init~ated. The following represents genegal observations about the form of the parses as it has emerged thus fax^.</Paragraph> <Paragraph position="8"> The rirst set of characteristics would result from the syntactic parsing of each definition. The purpose of this step would be simply to establish the syntactic pattern of each definltion. The output of this step would be similar to that generated by Winoqrad (1972) in h~s parser. The 'dictionaryt for the parser would be the very lictionary we are analyzing, although only the main entry, its inflectional forms, and its part-of-speech label would be used in this step. Ambiguous parsings and failures would be kicked out; the failures in particular, would provide an excellent source for refinlng the parser used by Winograd. Clewly, this step is fiat trivial, and, it might even be argued that it is beyond the state-of-the-art.</Paragraph> <Paragraph position="9"> However, by using a corpus as large as a dictimnary and by kicking out failures and ambiguities, I believe that this step will significant3 advance the state-of-the-art The second set of characteristics would be determined from a semantic parsing of the definitions, that is, an attempt to identify the cases snd semantic components present within each definition. For this study I have found the followinp; distinction to be useful: 4 case is a semantic entity which is not intrinsic to the meaning of a word, e.g. that eomeone ~s an agent of an action, whereas a component; 1s an intrinsic pert of the meaning, e.g. a human being is animate It is necessary to artl culate recognition rules for determining that a particular case or semantic component is present The 1ittA.e thdt has been done ta develop such rules has &en based prl~arily on syntwtic structures or a prlorl assertions that a given case or conponent is present. Despite the recc mized deficiencj es of dictionaries, I be3ieve that it is possible to-bring much greater rigor to such rules with evidence gleaned directly fi om the definitions. For example, - cut has a definition, &quot;penetrate with an ins%rurnenVt ; this defin~tion irJould be parsed as having the instrument case. (Note also that this definition makes the instrument case intrinsic to cut.) Havnver, in most cnnen. it will be necessary to examine the definitions of the ,constituent woras. For example, the verb knife has the definition, tf cut with a knifetf; although it is quite obvzous in this instance that a knife is an instrument, rigor demands that we go to its aef inltlons where we flnd, &quot; a sample instrument . . . &quot;. 4 great leal of analysls may ultimately be requlred to discern the intrash character~atics to be attached to a definition, but I beli-eve that many of these can come from the dictionary itself rather than grom ~ntuition.</Paragraph> <Paragraph position="10"> Although the nuaber of cases and components discussed in the literature is nut very large, the number of ways dn whlch they may be expressed, at least &n English, is slgnlficantly larger. In addition, there is sl ell a large amount of ambigutty, e , not every form spec.if.ically indicates the presence of a particular case. For example, a defjnaon, &quot;act - with haste&quot; does no$ indxcate that &quot;hasteft An an instrument: rather, &quot;with haste&quot; expresses a manner of actlng. Unraveling all these nuances requires a great deaE of effort. However, it appears that a partlcalarly good source of help i,n this endeavor might be found in the definitions of preposlti~ks (which are used pr3manly to indicate sense relations).</Paragraph> <Paragraph position="11"> Bennett 1975 found it possable to express the meaning of spatial and temporal pregosit:ions (a high percentage of all prepositions) with only 23 components. However, in Websterls.</Paragraph> <Paragraph position="12"> the number of then deflnltions is at least two ordera of mag nJ tudes hlgher. The d~fference seems to he in the &quot;apparenC polysemyu vrh~ch, as Bennett says, arlses from the lncluszon in preposl tl onal definltlons of &quot;redundant features already determined by the enylronment&quot;. In other words, many preposit,ional def~nltlons contam lnformatlon about the context surrounding the pr~osition, pa~ticularly what sort of entities are related by the prepositions. My examination of verb defintions containing prepositions haa led to the observation of many noticeable word patterns, i.e. collocations, which appear to be uaful xh the recognition ~f cases. For example, one definition of - af states that its object indicates something irom which a person or thing- is delivered&quot;. In examining verb definitions, there appears to be a distinct set of verbs with which tnls sense is used in the following frame ft( transit Sve verb) (ooject) of ( something)&quot;. The verbs that fit the slat are exemplified by free, clear, relieve, and - rld. Thus, if this pattern appears, the ob-ject of the preposition can be assigned the meaning It something from which a person or thing is blivered&quot;. cfChrough the use of prepositional bfinitions in this way, I have therefore been able to articulate some semantic recognition rules by which the arllst, or cage of a noun phrase the object c P a preposition) can be identified. My use of this technique has barely begun, so that it? is presently unclear whether this appmach will suffice to discldse all the caE informatl~n that we wish to identify ~ith a senantic parser, but if not it will tfertainly make significant strides toward this objective.</Paragraph> <Paragraph position="13"> Parsing of a definition according to the greceding notions is still not sufficient to identify'the semantic components which should be attached to a main entry, since much af the sermntic content is only present by virtde of -the definition's constituent wsrds. Thus, a compl ete rendering of a definnion' s semantic content must be derived from the sernantlc characte~istics of Its constituents, in a recursive fa~hiong all the way down to the primitives. Although -identification of these primitives is the primary go-1 of the aoproach being presented here, and Wce, intrinsically incomplete until the analysls is completed, the set ol semantic characteristics for a particular definitian can be developed as we proceed towdrd our goal. Yo do this, it will be necessary to articulate rules whlch indicate hou semantic characteristics may be transmitted from me definition to another. An example of such a rule is: If the noun X possesses the semantic component &quot;animatem, and if X iti the noun genus) the noun Y will also have the component ltanimate&quot;. Another exarhple is: If a verb X has a definition x whlch has been parsed as having an instrument case, and X is the core verb of a definitlor, y of Y, and y. also has been parsed as having the instrument case, J then the instrument in J is &quot;a type ofs tne i~istrurnent ia xi. j It will also be necessary to articulate other derivational (such as the application of a causative derivation to a state vqb) and transformational (such as the application of a rrerundid transformation to any verb) rules. This process of delineating how semantic characteristics are trmsmitted will at fhe same time give more meaning to the lines of the diotionary d~graph than simply &quot;is used to definett.</Paragraph> <Paragraph position="14"> The third, and final, set Q-f characteristics $hat must be attached to a definition is a s~ecifieation of the context that must be present if that def rn~tlon intended. The context re& stnc~lons may requme mat the deflnlendum must be used in a particular syntactical way, for example, as a trms~l lvn or intranslkive verb. Usage restrict&ons may speclfy the presence of partlcdar wor& such as particles or objects. For example, there is a distinct set of defin~tions for the ~dlom - take -.I out whrch thus requlres the presence of the partlcle &quot;outll zn addltion tc the verb. One definition of the transityve verb chuck requires the object Itbaseballt1. Other defln~tlons may requlre a speclflc subaect. ~~nilly, there are sernant1-e restrlttlons that may be dis~ernlble only from the definition itself. For examp3re two deflnltlons of the verb chew r re: &quot;to give new hope toit 7and llllft from olscouragement, dejection, or sadness to a more happy state&quot;; lf the seco~d deflnltlon 1s- uatended, it seems necessary that the context lndlcate the prlor state of dlscouragement, de jectlon, or sadness, slnce we cannot presume such a state, for someone mlght have been zin a happy or non-sad state dnd simply recelved some new hope. In the absence of the necessary context, we would default to the flrst defmxtlon.</Paragraph> <Paragraph position="15"> Thus far in my research, I have not devoted any effart toward, developln~ ~rocedures for prescrlblng the context based o* the deflnat~on. I expect that lnltlat~on of thls step'wlll beneflt f~pm further results of the first two steps.</Paragraph> <Paragraph position="16"> Although the parslng system outllned in thls section may appear to be exceedingly cornplaw, such an eventuality is not unex~~cted. The character1s&quot;t~os to be attached to each def in&30- null tion are not significantly different from those proposed by Fillmore 1971. It is also important to no-ce xnax some of the goals of analyzing the contents of a dictionary are to reduce the amount of redundancy, to remcnre vicious circles, ad to represent the meaning 6f a word in a more efficient way. Hopefully, this type of analysis would eventually leqd to a substantial reduction in the size of a dictionary; the prospects for thls are considered further in the next section, 8. THE ULTIMATE MODEL: POINTS AS CONCEPTS At this juncture, it is necessary to ask whether the points of the digraph models sufficiently corl espond to meaning as we wish it to be represented. In the two models described thus far, ,the analysis of a definition was deemed complete when the appropriate definitions of the const$tuent words had been identdfied. This situation 1s not entirely, satisfactory, since, if a constituent word has more than one definition that applies, the definitior being analyzed is subject to more than one interpre bation and hence may be called ambiguous with respect to that constituent. For example, if the two definrtions of abroad, &quot;over a wide -nrea&quot; an4 &quot;at largeH, fit the definition of broadcast to yield either Itthe act of spreading over a wide areat1 or &quot;the acd of spreading at larget1, it is not legitimate to exclude one. This situa Lion is only a reflection of the fact that natural language is almost always somewhat ambiguous. However, in accepting this fact, it is necessary that we incoqorate it into our models, Parts of the parsing system described in the last section will help to discriminate and select those defini-ti~ns of a constltuent word whioh fit n given context. As the parser is refined, the candidates for a particular context will be narrowed as described in Section 6, but many instances will remain where more than one del hition fits the context. We might say that any point representing more than one definition thus constitutes an ambiguity. Viewed differently, we might also scy that the context is not sufficient to distinguish among all the defitions of a word, In other words, we can-tbLamer the ambigu? ity on the context..</Paragraph> <Paragraph position="17"> We must expect that ambiguity will be present in the dictionary and deal with it on that basis. Fgr purposes of illustration, let us say that abroad shown in FLgure 4 1s one such point. To remove such points from the d graph, we must make two points for the definition oj broadcast, ope repsmenting &quot;the act of spreading abraaditt and one representing &quot;the act of spreading abroad2&quot;. These two points use the same words for expressin$ a definition and will-be distinguishable only by the fact that their underlying definitions are different. Because of this situation, it is no longer valid to say that a point of the model represents a definition: rather, we will say that a point represents a lfconceptfl.</Paragraph> <Paragraph position="18"> It is also pessibxe that the concepts represented by two or more points can be shown to be equivalent. Ihe concept, *'the act of spreading absoadft, has men shown to be equivalent to &quot;the act of spreading over a wide arealt. If the latter phraseology appears under some main entry, say distribution, then bath it and the definition of broadcast would eventually be an.</Paragraph> <Paragraph position="19"> slyzed in the same way. We will say that both expressions may represent the same concept and hence are equivalent at least to this extent. (since the-other definitions of these words would be dif fwent they are not totally equivalent.) This concept will thus be represented by one point, labeled by either - broadcast or distribution and equi'ra1ent.l~ verbalized as '!the act of spreading or &quot;the act of spreading over a wide This interpretation is a reflection of the fact that in ordinary speech a single coacept may be verbalized in mbre than one way, The observations in this sectlon lead to the following description of the 'ultimate' model: The semantic content of a dictionary may be represented by means of a digraph in which (1) a point represents a distinct concept, which may be verbalized in more than one way and may have more than one label, and to which is appended a set of syntactic, semantic, and usage features, and (2) a line represents an instance bf some one of a set of cxperators which act on the verbalizations or labels of a point according to the feafures of that point to ield the parametric values of another point. It should go without saying that the cpmplete portrayal of a dictionary according to this model requir,es a considerable amouht of further work; nonethejess, I believe that the model provides the appropriate- framework for describing a dictionary.</Paragraph> <Paragraph position="20"> 9. PROCBDURBS . FOR FINDING THE PRIMIEIVES In Section 3, I stated that the model of a dictionary shouid permi t Lhe transformation of each definition into i ta pqimitive- components. Based on the pneceding deacriptl ws. it is sugge~ted tha-c tHe 1x1~ articulatio~ of the ultimate model wxll satiqfy this objective for the following reasons: ( I) An elementary Fheorem in the theory of digraphs1 maerts that every digraph has a poin, basis, that is, a set of points from which every point in the digraph may be reached. Since points represent concepts in the ultimate model, it seems reasonable to assert that the point basis of its tligraph represents the set 09 prirnftive concepts out of which all bthers iri the dictionary may be formed. Based on the characteristics of the mints in that model, it is possible (and perhape even necessary) that each primitive cancept would be verbalized in several ways and symbolised in several way$ (as will be shown below) (2) Since the digraph has a finlte number of points and lines, the sets of primitive concepts and operators are also finite.</Paragraph> <Paragraph position="21"> It dllly remains Do ?find the primitive concepts; this will be done by applying rules, based oh the models and the parsing system, ta identify words and definitions which cannot %be primitives. Essentially, the assertion that a word or definition is non-prim~tive requires a showlng that it is derived from a more primitive concept and that a primitive cannut be derived from it. These non-primitives can be set aside and their Pull syntactic and semantic characterization can be accamplished after the primitives have been identified. Although no primitives have yet been identifieti (since the described procedures have not been fully applied), then form and nature will be delineated, null To dernoqstrate the validity of my approach, 1 have bPe~ applying rules developed thus far t'o the set of verbs in - Websterr's Thud New Internatronal Dictionary ( 20,000 verbs and their 111,000 definitions). This set was chosen because of their importance (cf. Chafe 19'70) and the (bare) feasibility of coping with them manually (although it may be another 3-4 years before I am finisheh,. at my current rate of progress). I have attempted to formulate my procedures with some rigor, keeping in mind the ultimqte necessity of computerieat~on. I have developed some detailed specifications for some of my procedures, envisioning the use of computer tapes developed by Olney, but have not completed these since I do not presently have acoess to a computer.</Paragraph> <Paragraph position="22"> Despite the focus on verbs, it will become clear that words from other nnrts of speech are inextricably involved in the analysis. Also, the rules that are presented can, for the most part. be applzed to other parts of speech. lotwithst;indi?lg the fact that the meaning of many verbs is derived in part from nouns and adjectives, I believe that each verb definition alsu, contains a primitive verb constituent.</Paragraph> <Paragraph position="23"> Lacn vero aeunltlon conslsts of a core verb (~bllgatory) and some dlfferenxlae (opt~onal). he deflnxtions of other parts of speech have a similar structure, i.t. a core unit fron the same part of speech and some hfferentlae.) The subgraph of the total dictionary digraph formed by core verbs accords fully wlth the models described LJI Sectlane 4, 5, and 7. Therefore, any rules developed on the basls of those models wlll apply equally to the verb subgraph. We need only keep An m~nd that the differentiae come from other parts, of speech and become embodied ~n the core verb. Thrs 1s Bow the verb - cu% comes to have the lnstrurnent case ~ntrinslcally. To begln the analysis, we will let E represent the set of those vnrb deflnltlons whlch have been Adentifled as non-prlmxtlve; ~nltlally, thls set i& empty.</Paragraph> <Paragraph position="24"> Rule 1'. 4f a verb maln entry is not used aa the core unlt</Paragraph> <Paragraph position="26"> of any verb definltlon in the axctlonary,' then all its defznltlone-may be placed in B. (Thas rule applles to points of the baslc model whlch have outdegreq 0, 1. e. no outgolng ilnes. ) Slnce no points can be reached \from such a verb, ~t cannot be Flgure 5. Basic model, verb subgraph example subject to Rule 1.</Paragraph> <Paragraph position="27"> primltlve. 131 Flgure 5, the pornt labeled by pram represents -the defin~tlon Itto air (as a chlld) In or as if in a baby carr1ageff ; slnce pram is the core unlt for no definition in the dictionary, all its definitions mav be excluded as non-primitive. In W3, this rule applies to approximately 13,800 verbs out of 20,000; the number of definitions in the verbs excluded is not known, Rule 2. If a verb main entry is used only as the core unit of definitions already placed in E, then all its definitions may also be placed in E. his rule applies to points of the basic mceel with pusitive outdegree.</Paragraph> <Paragraph position="28"> The uses of su* verbs as core units follow definitional paths that dead-end; hence, they cannot be primitive. Figure 6 shows a portion of the dictionary cover cake barkle rkgure 6. Basic model, verb subgraph examplie subject to aule 2.</Paragraph> <Paragraph position="29"> digrapn where the verb - cake defines only barkle, which in turn is not used to define! any verb. Thus, the definitions of - cake may be included in E after the definitions of barkle have been entbred, In W3, this rule applies to approximately 1400 of the strong componwt or in definitions of verbs already placed in E by Rules 1. 2, or 3, =then the definit,ions of all verbs in the strong component may be placed in E. (This rule applies to points of &he basic model which constitute a ptrong component, i. e. a maximal let of points such that for every two points, u and v, there are paths from u to v and from v to. u.</Paragraph> <Paragraph position="30"> This rule does not apply when EUhe strong component consists of all points not yet placed in E.) 4 strong component consiqtAng of the verbs aerate, aerifg, air, and ventillate is shown in Pimre 7. Except for oxygenate, the other verbs defining the set constltuting the strong component are not shown. Shce it is possible to start at any of the four and follow a path to any other of the four, there as no real generic hierarchy among *them. It is possible to emerge from the strorlg component and follow paths to pram, eventilate and perflate, to whlch, however, Rule 1 applies. If we follow a definitional path that leads ihto thls strong component, we can never get out agaln or if .we tlo we will only dead-end. Hence, the de finitions of all the verbs in the strong component are not primitive and may be placed in E. In WJ, this rule applies to approximatelv 150 of the 4800 remaining after the application of Rule 2. Actually, Rules 2 and after Rule 3 places the%def~nltlons of aerate, aer~fy, -9 alr and. ventilate in b, it so happens that ilule 2 then applles to the definitions ~f oxuenate.</Paragraph> <Paragraph position="31"> After Rules 1.2, and 3 are applied t~ the digraph or the baslc model, tne remaining polnts constitute a strong component of approximately 4500 polnts. Thls dlffers from those to which Rule 3 applies in that there'would be no ~olnts left if we placed all it8 polnts in E. phis flnal stro~g component 1s the basls set of the basic model, that is, any point of the basic model (1. e. any main entry in the dictionary) may be reached from any point in the final strong compo'nent (but not conversely) * kt th.is juncture, we can proceed no further w.i;h the basic model alone; it .is necessary to expand the points of the final strong component lnto two or more points each representing a subset of the definitions represented by the orlglnal point, as previously shown in Flgure 3, In part, this can be laccompllshed by ~derl,Llfylng ~ndividual definitions which are not used. Rule 4. If !any definition can be shown to be not used as the sense of any core unlt (or only those already in E), it may be placed in E. Th,is rule is essent:ia$ly a restatement of Rule 1 for xad~vidual definitions and includes the following two subrules, among others nat presented.</Paragraph> <Paragraph position="32"> Rule 4a. If all the rema,in,ing uses of a verb are trans1 tlver ( intransitive) then .its :intransitive ( trans.it.ive) defini* null tions are not ,used and may be placed in E. The exp'anslon of a poinl into transitive and intransitifire uses is a good examole of how the points of the basic model are transformed into pgints of the expanded model.</Paragraph> <Paragraph position="33"> Rule 4b. 12 a definition is rn ked by a status, label (e.g. archaic or obsolete),, a subject labs or a subject-guide phrase, it may be pldced in E. Lexicographers creating W3 were instructed not to use such marked definitionn in defining any other word. .</Paragraph> <Paragraph position="34"> Other ?rules have been developed in an at,kernot to identify $he specific sense of the core verb, or those senses of n verb which have not been used in deyining other verbs, but are not presented here. However, there are too many instances where the differentiae of a definition do not provide sufficient context to exclude all but one sense (for example, many senses of - move fit into a definition phrased &quot;move quickly&quot;). In order to continue toward the primitxves, we must shift gears slightly and ask whether a definition can be characterized as llcomplexv, that is, derived from more primitive elements. For example, one befinition - of - make is &quot;cause to be&quot;, which can be labeled as complex aecause it conbists of a causative, component and a state component, each of which is more primitive by it~elf than lcause to be&quot;, The importaqqe of the notion of a complex definition becomes evident when we try to viaualiee how a primitive concept wd.11~ be identified. To understand this, -re must consid-er some further properties of the digraph. After the application of nuie and any Subsequent rule), the remaining graph is a findl sProng component. (~ecall that in a strong component, for each two points, u and v, there is a path from u to v and one from v to u. ) Assuming that each point aepresents a concept (as in the ultimate model), the fact that two concepts are in the same strong component means thdt they are e~uivalen4. In more traditional terms, what we have is a definitional vicious circle, . that is, a definitronal chain which adds nothing to our Undeistanding of the meahings invol ved.</Paragraph> <Paragraph position="35"> Using the digraph of the final strong component, we can identify (and examine one by one) all putative definitional cycles or vicious clrcles; these will fall into three classes. The first class will konsist of improper cycles, whiqh can be removed by determining that one poifit is more complex (and hence not equivalent to the definition from which it is derived) Further rules for_ characterizing a definition as complex are given below. The second class of cycles will be real viclous circles, which fortunately can be removed, but only under certain conditions. For example, one definition of jockey is &quot;maneuver for advantage&quot;, while one definition of maneuver is &quot;jockey for positionN; these two definitions constitute a vicious circle. In order to remove it, the~e must be some other definition of either verb whicn constitutes its meaning; in this case, it is found under maneuver, specifically, &quot;shift tacticst1. Thus, in order to remove a vicious circle, we must find some way out. If we cannot, we have the third class of cycles; this class will comprise %he set of basic concepts. If there had been no way out for the example of jockey and maneuver we would have said that no meaning was conveyed by eitber Vera, but rather that the meaning was established by use. This third set of cycles is what is sought by the procedures described inathis paper.</Paragraph> <Paragraph position="36"> As mentioned above, the crux of the analysis after the application, ~f Rules 1 to 4 is the iaenfjfication of complex con& cepts.-Essentially this entails a showing that, for any definitlon yi of verb Y, with Y as the core verb of definition x of verb X, the differentiae of x, make yi generic to x . For exam-</Paragraph> </Section> <Section position="14" start_page="38" end_page="48" type="metho"> <SectionTitle> 3 J </SectionTitle> <Paragraph position="0"> ple, all transitive definitions of - cut would be generic to a definition in which &quot;cut1f is used with an object, even without narrowing down to one definition. The general rule may now be stated, Ru1.e 5. If any definition is identified as complex, it may be placed in E, The net effect of this rule is to brea~ one or</Paragraph> <Paragraph position="2"> more putative cycles hf equivalent definitions or concepts, enabling them to be transformed into a strict hierarchical order which will eventually be subject to Rule 4. Thus, the complex defiriition and all definitions that can be shown to be derived therefrom dan be placed in E, be cause they cannot be part of a primitive cycle.</Paragraph> <Paragraph position="3"> Rule 5 is implemenBed only by very specific recognition rules, which are essentially part of the parser. The specific rules entail a showing that some component has teen added in the differentiae of a definition that is not present in the II meanings of its core verb. For example, the limannern component is not htrinsic to the meaning of the verb moveo therefore, when a daiinition has the core verb flmovett with an adverb of manner, it can be marked as complex. In establishing a component as non-intrinsic, it is necessary to articulate rules for recognizing the presence of the &quot;mannerft component (such as a phrase in.8 manner&quot; or an &quot;-ly1! word with a aefinition 'hn a mannerl1) apd then to deterrnlne if that component is present in any definitions of a particular verb. If not, then the verb can be labeled as complex whenever it is used asthe core verb in a definition with differentiae that% fit the recognitinn rule. In addition to move I have determined that, for the manner component, the verbs act , perform, ater, speak, exII null press, behave, and many others follow the rul e. Table 1, on the next page, identifies some specific components, a brief descxiption of how they are recognized, some of the verbs to which the particular rule applies, and an -example of a gefinition labeled as complex by the rule and hence placed In E, If a definition has a. ccre verb whose applicable sense is one which has been marked as complex, it too can oe so marked, since it is derived from a complex definition. For example, all definitions of the forq &quot;make aajectiveqi, i.e. with an adjective complement, are deri'ved from the definition of -9 make &quot;cause to be or become&quot; and hence can be marked as complex. Tn add-ition, if all defini ions of a verb have been marked as</Paragraph> <Section position="1" start_page="42" end_page="48" type="sub_section"> <SectionTitle> Recognition Rules for Sernant~c Components Name of Examples of </SectionTitle> <Paragraph position="0"> Component Recognltlon Rule ApplFcable Verbs 1. Definitions Verb t cease, bean, commence vi -, 2, ~nf tnht$ve strive, continue &quot;begin to bi;&quot; 2. Causative ~ausat;ive verb cause, force, confront vt 2a, + Inf in1 tive compel, induce compel (a peraon) to face, fake account of, or enQuyett pake vt IOa, Vause to be 1. Instrument Verb t ttwithlt apply, fasten, knife vt 2a, + noun defined cut, beat l1 cut with a as instruwnt, device, etc.</Paragraph> <Paragraph position="1"> 4.Means Verb + &quot;by&quot; t make, prepare, draw vt 4e4,, (Process) Gerund form, shape l1 shape (glass) -.</Paragraph> <Paragraph position="2"> by drawing molten glass from the furnace over a senes of automatic rbllerstl 5. State Entry Verb + ~hnton + br rngi put, noun defined throw. fall as &quot;the state of ,,,</Paragraph> <Paragraph position="4"> compleq then all definitions in which it appears as a'core verb can be similarly marked and placed in E.</Paragraph> <Paragraph position="5"> Through the devel~pment and application of further papsing rules under Rule 5, I am hopeful that I will eventually arrive at the set of primitive verb concepts (i.e. cycles or vlcious circles with no way out). I have already reduced the number of verbs from 20,000 to less than 4,000. This number would be mucH lover, But TOY the fact that I am applyiqg the rules manually and I must ex8yiise %me-consuming,care to emure correatmess.</Paragraph> <Paragraph position="6"> After the primitive concepts have been identified, it will be necessary to gg back to all the definitions that were set aside in the process of finding the primitives, so that then semantic characteristics can be articulated. 5 fully expect that the parsing system which will have been deveoped will be able to accomplish much of this task I also expect that the parsing system will have equal applicability as a general parser capable of formally characterizing ordinary discourse in a canonical form. Of course, verification of this expectation will have tn await a full presentation of the parser.</Paragraph> <Paragraph position="7"> 10. R8LATIONSHlP TO'EFFORTS TO REPRESENT KNOWLEDGE IN FRAMES The process whkch has been outlined ifi the preceding sections is closely akin to current efrorts to represent knowledge in frames. (~f. Winston 1977 for an elementary presentation* of 'this notion.) Briefly, a frame consists of a fixed set of arguments, some of which may be specif.ically related to others, and some of which may have specific values.</Paragraph> <Paragraph position="8"> frame 1s intended to rppresent a stereotyped situation, with the arguinems identifying the various attributes which the situation always possesses. In terms of case grammar, for e~ample, a movement frame will contain arguments or slots for an agent, an instrument, and a destination.- By tying frames togeSher in spepific relationships, we can build larger and larger frames to represent more and more knowledge, perhaps constructing a series of events, an inferencp structrye, or a de8c~4ption of a scarre.</Paragraph> <Paragraph position="9"> Before bnilding these large structures, it is necessary to represent very small pieces of knowledge. Heretbforc, this has been done by postulating the components of frames to represent such things as actions and state changes. But- this can be accomplished an a more rigorous basis. Por example, if we first locate all definitions using &quot;move&quot; as its core verb and then identify all the case structures in which it 1s used, we wqll have a generalized frame which characterjzes most if not all of the possible uses of *lrnove1l. (This approach ds currently being followed by Slmmons 1977.) Each definition in which tlmovell is used could then be representea by the generalized frame with some of itb slots fllled. This process can be followed for any word for which we wish to develop a frame, If ,_ ln addition, we analyzed the definitions of -* move we will find that they, in turn, represent instantiations of still o3her frames, which will be even more generalised than those developed for the uses of ftmoveft, The difference between the frames representing the definitipns of - move and those represent- null ing the uses of &quot;move&quot; is that the latter are the same as the former wlth some slots filled. Within the bounds of the ambl~uity preselit in the dictionary, this stut-filling will identlfy which definition of - move are employed in which uses of umove&quot;. It seema C.0 me that this ir nothing ,more than the process which has already been described using a graph-theoretic naproach, except that the generalhxed frame for each verb will not be, carzied along tnrough each step. Moreover, si.nce the semantic parslng sy-stem which has been described wi13 be based largely on the relationships derived from EUhe definitions of prepositions,, , and these comprise most of the case relationships, the parsing system will effectively circumscribe the ~errnissible elements (i. e. slots) which can be, present, glven any particular. context. Thus, although the phraseol~y ik different, the effect is the same, If there is ali essential equivalence between these two approaches, then. since frames purport to represent knowledge, the process described; if successful, will result in an articulation of whatever knswledge is contained in a dictionary, What this implles is that the lexicon cofitains a great deal of knowledge about the world and not just information which will er+ able us to understand such knowledge, Frames provide a 'great deal of inslght to the approach which has been described here, but the reverse also seems to hold true. If the semantic content of each defihition can be captured, then it map be possible to~articulate the frame for any utterance by combinihg the characteristics of the definition$ of Yts constituent words within what\is permitted by the parsing system.</Paragraph> <Paragraph position="11"> In Section 1, I described some limitatzons of this paper and my research. This paper suffers from a lack of sufficient detail to enabl,e a reader oP researcher to replicate what I have done or to take the next steps of cbmputerizing the procedures whrch I have developed. I will provide further details bn the specific steps I have followed in reducing the set of verbs from 20,000 to 4,000 to anyone requesting. With respect to compqter specifications, I have prepared some, but stopped because I have no access to a computer, However, if any researcher 1s interested in pursuing this (or setting graduate students to work), I am prepared to develop the necessary specifications and to work hand-in-hand for the further advancement and refinemekt of this methodology, I also ~ndicated in Section 1 that my research presently shows no final res,z4lts and that I do not even know how much further effort-will be necessary to explicate tfi% parsing system which has been described. Clearly, there are great dmtances yet to be covered toward a goal of being capable of transformin& ordinary discourse into a canohical form. I believe that characterization of the contents of an ordinary dictionary If it seems worthwhile to pursue this approach, despite the limitations, I believe the best way to do so would be to establish a single computer-based repository for a dictionary, preferably W3, with @h-line access to researchers across the count-... LJ, and to build the parser and definitional. charac-terizations piece by piece. (I have noted how the parsing system which I have described can be built incrementally.) The magnitude of this effort Precludes much progress by individual researchers.</Paragraph> <Paragraph position="12"> Olney tried to do something similar with the collegiate dictionary baged on W3, but by distributing bulky computer tapes. He was unfortunately premature; it may be that now is th? time to try again,</Paragraph> </Section> </Section> <Section position="15" start_page="48" end_page="48" type="metho"> <SectionTitle> PRESIDENT TO SUPPORT LIMITED PRIVACY INgTIATIVE </SectionTitle> <Paragraph position="0"> Consistent wPth the Lelectlve approacn of the U S to privacy ~egulatlon (versus the onmibus app~oach of the Europeans on the subject), the Carter Ablnlstxatlon is expected to support & hmited legislative program in the 96th Congress on privacy issues.</Paragraph> <Paragraph position="1"> The President's response to the recommendations of the Privacy Protection Study Commission and previous legaslative efforts, termed the priv y initiative, i$ energlng r from a year-long study by an ad hoe group wit in Mr. Carter's Domestic Pollcy Staff. The study died &quot;Baby Blueff (compared with a largd , supporting b lue-colored doment called &quot;Big Blueu) was delivered to the President last December. The group, known afthe White House Brlvacy Study Coordinating Cornmitree, 1s headed by Stuart E Eizenstat, Assistant to the President for Domestlc Affairs, and Juanita M Kreps, Secretary of Commerce.</Paragraph> <Paragraph position="2"> Atlmlnistration Proposals Jt 4s reported that Mr. Carter may mention the prlvacy initiative In Rls State of the Union Address in January The Adminxstrat ion' s proposds are expected to r;ent er oh lamit ing Federal access to data in the prtvate sector, z,e., in the area of medicine, credit and insurance. The Privacy Coordinatu~g Committee recommended that these limlts on access should apply equally, to state'and local governments The Committee endorsed Federal lqp~lat ion leav~ng st at es to adopt laws Ifthat meet certain mrnimum standartis. &quot; The. privacy proposals would give individuals the rlght of &quot;ownership&quot; to personal data maintained in ap medical, credit and insurance sectors r ,-</Paragraph> </Section> <Section position="16" start_page="48" end_page="48" type="metho"> <SectionTitle> IN THIS ISSUE PFESEGIDENT TO SUPPORT LIMI~~D PRIVAGT INITIATIVE </SectionTitle> <Paragraph position="0"/> </Section> <Section position="17" start_page="48" end_page="76" type="metho"> <SectionTitle> AFIPS IN WASHINGTON WITNESS STATEMENTS AVAILABLE THROUGH WASHINGTON OFF ICE DETAILED </SectionTitle> <Paragraph position="0"> Thus, individuals would be entitled to review information in order $0 correct errors. (Aetna Life G Casualty Co. has initiated a similar policy,' at the urging of William 0. Bailw, Aetna Life president, and former Privacy Protection Study Commissian member.) It is possible that this right of Mournership&quot; will be incorporated into legislation amending the Fair Beds* ~apor.t;{ng Act. The proposals would also forbid disclosure of information where there is an expectation of ~onfidentiality.~~ The Gonupittee agreed to exclude a recommendation that wnllld encompass computerized telephone records. The Administration ' s privacy age~da seems to coincide with that of Rep. ~icfiardso~ Preyer (D-N. C.) who predicts the Congress will consider measures concerning medical records, banking recbrds and third-party records.</Paragraph> <Paragraph position="1"> 'Administrative Steps. Besides the legislative proposals on privacy, the President- is expected to take some l1adm$nistrat ive steps, tt using executive authoriza kion (see WashCngton Report, 12/ 78, p. 11).</Paragraph> <Paragraph position="2"> 'International Information Issues.' The privacy initiative precedes exp-ted fi Pture Administration proposais on so-calded &quot;mternational Information issues, I' such as overseas restrictions on transborder data flow, the transmission of data across international boundaries. Henry Geller, Assistant Secretary of Commerce for Communications E Information, has noted it is time for the U.S. to l5bring' . . . [~ts own] house in ordert1 on privacy issues (see Washingtm Report, 12/78, p. 11).</Paragraph> <Paragraph position="3"> Role of the Computer. Rec~gnizing the role of the computer in facilitaring the collection and dissemination of information, Carter official% state that legal protection against the indiscriminate use of data has not developed as rapidly as the technology. In one draft of the report prepared for the President by the Privacy Coordinatilig Committee, the group noted that, &quot;We are faced by a slow but steady erosion of privacy which ~f left unreversed, will take us (in another generation) -to a position wherce the extent of our human rlghts and vitality of our democracy will be j e~pardized . &quot; Previous Privacy Legisl atlon. The Pres identlal Prrvacy init lati-ve follows passage of the Privacy Act of 1974 and the Right to FinrmciaZ * Privacy Act (FJashington Report, 12/78, p. 1) . The Privacq Act Limits Federal agencies acceks to personal informat ion held by other Fadera1 agencies. The Right tt~ ~inaricia~' Privacy Act limits Federal access to personal lnformation in the fin&clal sector. Cited as ~majd?*achlevement by the Carterr Administration, the Financia2 Privacy Act has been criticjzed by certain individuals for increasing the potential number of bank examinations conducted by Federal investigators; for lacking sufficient legal grounds to challenge unreasonable access to data; and for exempting political action groqs. [An internal audit, made public recently by the U.S. Postal Service criticizes the Post Off ice for inadequate implementation of the %u&y Act of 1974.1 Effect of Gongressional Elections on Privacy Issues. The surprise defeat of Rep. Edward W. Pattison ID-N.Y.) in the November Congressional elections removes a staunch defender of financial privacy legislation from the House ~dking Committee. Also, on the SsRate side, Sen. Thomas J. McIntyrets (D-N.H.) loss is expe~ted -to+change rne character of the FEBRUARY, 1979 2 AFIPS WASHINGTON REPORT Financial Institutians Subcommittee which the Senator chaired. koweyer, 77 strong privacy advocates were elected to the House of Representatives in California: a Democrat, Vic Fazxo, sponsor of a Fair Information Practice Bill enacted in California in 1977; and a Republican, Jerry Lewis (no relation to the pnt-ertainerl, spoqsor of additional. state-wide</Paragraph> </Section> <Section position="18" start_page="76" end_page="76" type="metho"> <SectionTitle> AFIPS IN WASHINGTON WITNESS STATEMENTS AVAILABLE 'I%ROUGM/WASHINGTON OFFICE DETAI LED </SectionTitle> <Paragraph position="0"> The AFIPS Washington Office has compiled numerous-witness statements made before the Executive pd Legislative Branches of vvernment on information policy issues as part of a Witqess Statement Exchange initiated last year (Washing,ton Report,lll/78, p. 6). For participants in the witness statement exchange (rules for participation 'dbscribed below), the $bllowing wltness statements may be obtained s H. R. 214, The Bit2 of Rights Procedms Act. Philip B. Heyman, appearing ~uf~ 13, 1978, before the House Subcommittee on Courts; Richard J. Davis, Assistant Secretary of the Treasury, Enforcement and Operation, 3epartrngnt of the Treasury, appeariGg July 20; - 1978; and Paul 'G. ~oe, ~ssistant Chief Pa jtal Inspector, -CriminqA Investigations, U.S. Postal Service, appearing July 20th.</Paragraph> <Paragraph position="1"> H.R. 13015, 171s Conununications Act of 1078. Tyrone Brown, commissioner, Federal Communications Commission (FCC), appearing July 18, 1978, before the House Subcommittee on ~ommunications; krgitaW E. hite, commissioner, FCC, appearirig July 18, 1978; James H. Quello, commissioner, FCC, appearing July 18th; Philip S. Nyborg , vice-p9esident &id general counsel Camputer</Paragraph> </Section> <Section position="19" start_page="76" end_page="76" type="metho"> <SectionTitle> 6 Communications Industry Association (CCIA) , appearing August 3, 1978; </SectionTitle> <Paragraph position="0"> Charles b. Ferris, chairman FCC, appearing ~u~ust 9, 1978; Joseph R. Fag-ty, commissioner, FCC, appearing August 9th; MaTgita E. White, conunissioner,' FE, appearing August 9th; L. C. Whitney, president, National Data Corp. , appearing August 10, 19J8; and Herbert N. . Jasper, executive vice president, Aa Hoc Committee for Competitive Telec~mmunications. appearing August lWth.</Paragraph> <Paragraph position="1"> S. 2096, The 'Right to Financial Privacy Act of 1977, and S. 2293, The EZsotronio Fwd8 Tmsfer Act of 1977. Robert Ellis Sgith, publisher, Privacy JournaZ, appearing May 19, 1978 ,Tefore the Senaxe Subcommittee on Financial Institutions.</Paragraph> <Paragraph position="2"> S. 3270, The Jus*ice. System Improvsment Act of 2978. Jeffrey A. Roth, senior economic analyst, Institute for Law & Social Research, appearing ~u~ust-23, 1978, before the ~enate,~ubcommitt~e on-Criminal Laws 6 Procedures; also, James E rfke Cameron, chairman, Conference of Chief Justices, appearing August 23, 1978; Patrick V. Murphy, president, Police Foundation, appearing August 23rd; and Glen D. King, executive director, Internatiohal Association of Chiefs of Police, appearing August 23rd.</Paragraph> <Paragraph position="3"> FEBRUARY, 1979 AF IPS* WASHINGTON REPORT</Paragraph> <Section position="1" start_page="76" end_page="76" type="sub_section"> <SectionTitle> ' Confidentiality of Medical Records. ' </SectionTitle> <Paragraph position="0"> Richard I. Beattie, deputy general counsel, Department of Health, Education 6 Welfare, appearing May 23, 1978, bsfote tWe Hwse Qybcommittee on Government Information 6 IndiVidual Rights.</Paragraph> <Paragraph position="1"> EX arts.! Juanita M. Kreps, Secretary of Commerce, appearing September 28, efore the Senate Committee on Commerce, Science 4 Transportation. * 'Future of !hall Business in America.' John H Shenefield, assistant attorney general, Ahtitrust Divtsion, Department of Justice, appearing July 20, 197,8, befoae the House Subcommittee on Antitrust: Consumers & Employment; and d. G. W. Bi idle, flesident, CCIA, appearing July 20, 1978. High Technol'ogy Businesses. Jean N. Tariot , chairman, Incotem COT. , appearing July 20, 1978, befqre the Joint Senate Committee on Small ~usiness and House ~ubcommit tee on Antitrust, Consumers 6 Employment ; and Lester A. Fettigj administrat&, Federal ~rocuremek Policy, Office of Management 6 Budget, appearing August 10, 1978.</Paragraph> <Paragraph position="2"> Rules for Participation. To participate in the exchange of statements made before the Executive and Legislative Branches of Government on informagion issues, one recent witness statement concerning informasion policy should be ssnt to: Pender M. McCartex,' Research Associate, MIPS Washinaton Office, 1815 North Lyqn Stbet, Suite 805, Arlington, Virginia 22209. Thus enrolled in the program, Specific witness statement reqiiests gan be made (based on the above list), by mail only, enclosing a ,stamped, self-addressed envelope. For each requested kitness stat-nt, one statement shwld be included, in add'ition to the 'first establ5shing participation in the proera. It is not necessary to be a witness in a hearing; having access to such statements is sufficient. Updated listings of available witness statements will be issued periodically.</Paragraph> </Section> </Section> <Section position="20" start_page="76" end_page="81" type="metho"> <SectionTitle> SPECIAL REPORT </SectionTitle> <Paragraph position="0"> are involved in kuropean restrictions on the transmissio~ of data across ~nternational boundaries. (The CE is preparing a 1980 treaty concernitlg transborder data flow. ) ~ocord:din~ to M. Hondius, such bod-ies as the 20-member CE (in which the united States is oay a non-voting member) are seeking to protect lpeoplets rights and interests.&quot; He added that the European goal is to &quot;protect people against computers and computers computer technology on human terms. &quot; Mr. Freese added that it was his philosophy to &quot;try to solve proble~as before they occur. &quot; General Principles of Data Protect ior, Cited. Hondius out lined some general principles of data protection laws already in effect in some seven countries. (~~~roxirnatel~ seven more nations are expected to follow these countries with their own privacy legislation. ) The three principles are: (1) Publicity: &quot;People should know what is going on in general&quot;; -(2) Propriety: ''Data systems should be proper&quot;; and (3) qontrol : &quot;Recordkeeping should observe norms. &quot; .</Paragraph> <Paragraph position="1"> U.S. Privacy Policy Criticized. While stating that U. S. laws such as the Privacy Act of 1974 did represent &quot;a legislative step forward,&quot; Professor David F . Linowes, former chairman, Privacy protect ion Study Commission, said that the Pritracy Act provides &quot;no benefits PS0- the general publict'; coqtains too many exceptions and tuoOfew penalties; and disregards accountability.</Paragraph> <Paragraph position="2"> Computer users from large mu1 tinat ional corporations at tending the conference criticized the U.S. for a lack of leadership in formulating a position on issues involved in transborder data flow.</Paragraph> <Paragraph position="3"> According to ane FEBRUARY, 19 79 5 AFI PS WASHINGTON REPORT PROF. LINOWES, POLITICAL ECONOMY 6 PUBLIC POLICY, UNIVERSITY OF ILLINOIS (AFIPS/ P. Y. McCarter) account of &quot;an informal, not-for attribution meeting,&quot; held after one of the conference sessions, the users formed an ad hoe committec to lobby on transborder data flow issues.</Paragraph> <Paragraph position="4"> U.S. Industry Criticized. Administration officials appearing at the conference reiterated their criticism of industry for not becoming involved in the issues, and implored industry to provide specific instances of economic harm caused by restrictions on transborder data flow. Attending the conference and named as primary contacts for industry were: William Fishman, deputy associate adminifirator for Policy Analysis and Development, National Telecommunications C Infor~ltion Administration (NTIA), U.S. Department of Commerce; and Morris - H, Crawf ord, Bureau of Oceans G 1nter;ational Environmental ti Scientific Affairs, U. S. Department of State.</Paragraph> <Paragraph position="5"> OhCD Drafting Group Meeting Ileld. The Drafting Grow of the Organization of Economic Cooperation G Development (OECD) met December 6-8 in Paris to consider a new draft of ~ransborder Data Flow Guidelines prepared by Peter Seipel, consultant to the OECD Secretariat ( WashCngton Report, January, 1979, p. 1) . Attending the meeting as U. S. representatives were : L~cy ~wnmer, Esq., ~e~artment of state; Will iam Fishman, NTIA; and James Howard, NTIA.</Paragraph> <Paragraph position="6"> Inclusion of Manual Files, 'Legal Persons' Debated. At the OECD meeting, there was substantial disagreement on including manual files as well as computer files in the draft guidelines. In addition, the delegations were divided on extending privacy protection to lllegal persons&quot; (Le., business FEBRUARY, 1979 6 AF I PS WASHINGTON REPORT corporat iong and various other organizations) as well as individuals. The Europeans favor a more comprehensive approach to privacy legislation and generally view as ineffectual the selective approach taken by the U.S. Consensus Said to be Supporting U.S. Position. Despite these recent developments, a consensus is said to be growing in both the OECD and the Council of Europe supporting the U.S. position. For example, the latest Seipel draft has been interpreted by an Administrati on source as being &quot;very favorable&quot; to the U. S. position.</Paragraph> </Section> <Section position="21" start_page="81" end_page="81" type="metho"> <SectionTitle> NEWS BRIEFS </SectionTitle> <Paragraph position="0"> A recommendation for a 3ecial Assistant to the President for Information g. 51, has been&quot; dropped in a final draft; according to the most recent version of the consensus report [now circulating amopg Cabinet and Office of Management 6 Budget (OMB) officials], the FDPRP majority view &quot;holds that the . . . [FDPRPIreqommendation - can and - must be implemented through a strong and persevering Presidential injtlatlve through the OMB. . . .I1; the OMB is expected to pqesent the consensus report to the' President after final revisions.</Paragraph> <Paragraph position="1"> A formal study &quot;to determine the Administration's policy . . , [on] the future role of the U.S. Postal Service in providing services by electronic ~omrnunications~' is bei~g initiated by the White House under Stuart Eizenstat. the Assistant to the President for Domestic Policy; an Interagency Coordinating Committee, chaired by Mr. Ei zenstat , met December 13th to outline electronic communicationst issues; the National Telecommunications E Information Administration, designated as &quot;lead staff agency&quot; for the study, is soliciting comments from &quot;intergsted individuals or organizations&quot; to be considered in thp development of the Adminlstrationl s position; Congress is expected to address the issue this Spring.</Paragraph> <Paragraph position="2"> In December, the Postal Service Buard of Governors authorized temporary implementation of E-COM service, an electronic message service (EMS) for large-volume users (see Washington Report, 11/78, -p. 3); in November, Postmaster General William F. Bolger approved a four hllion dollar electronic mail experiment beginning this year; also in November, Xerox COT. filed a request with the Federal Conununications Commission to reallocate a portion of tbe radio spectrum for EMS.</Paragraph> <Paragraph position="3"> [Bletter information- is needed . . . to make assessment and evaluation of the policy alternatives regarding CCH [the computerized criminal history file],&quot; according' to an Office of Techno~ogy Assessment (OTA) study released in January, the first phase of a new OTA assessment of the Social Implications of National Information systems ; entit led A Pre timinary AssessmentL of the NationaZ Crime Information Center and the Computerized CriminaZ History System (#- -endose $2.75) , the study notes, &quot;Although CM has been the subject of numerous studies, conferences FEBRUARY, 1979 7 AF I PS WASHINGTON REPORT and hearings, there is only limited information regarding the ways is which law enf~rcment and the criminal justice decisionmakers, as well as other government and private individuals and the press make use of criminal history information, its benefits, the value of nationwide access to information, and the value of rapid access.</Paragraph> <Paragraph position="4"> The General Accounting Office (GAO) is preparing to release a new study entitled Security of Automated Information Systems of Federa2 Agencies; according to ia tentative outline of the GAO report, obtained by the AFIPS Washington Off ice, I1organizatiwnal structures1\ are I1inadequatelt and lfcomprehensive procedures&quot; are nonexistent in current Federal security precautions.</Paragraph> <Paragraph position="5"> A research and development project to evaluate the use of data encryption devices in protecting the Federal Reserve System1 s, (FRS) Fedwire operations is expected to be completed this June; Fedwire, a form of electronic funds transfer, links FRS to member ganks nationwide.</Paragraph> <Paragraph position="6"> In December, the Department of Justice said it is considering computer crimc involved in counterfeit or stolen securities as well as bribery and kickbacks.</Paragraph> <Paragraph position="7"> The Federal Cofimunications Commission - (FCC) 1s expected to add the Corrputer Inquimj II to its weekly agenda again, after two previous postponements; the FCC may determine whether AI'ET, a regulated communications common carrier, can provide unregulated data processing sefvices.</Paragraph> <Paragraph position="8"> The Supreme Court is eonsidering whether, under the Freedom of&quot; Infomation Act, individuals &quot;can obtain confidential business data; in November, the High Court let stand a U.S. Court of Appeals decision (Washington Report, 6/78, p. 4) allowing MCI Communications Corp. to use AT6T1 s local phone conn'ection to impleme,nt Execunet , hlCI1 s long distance telephone service providing voice and data communications.</Paragraph> <Paragraph position="9"> In December, the Office of Management Fr Budget (OMBI issued for comment a directive which would require Federal agency data processing users to account for the future cost of their DP systems; also in December, It OMB issued an annotateh bibliography (#) of current laws, panlcies, regulations, and &quot;guidance d,,cumentslV which are relevant to the acquisition, mqnagement , ana use of Federal data processing and related telecommunications resources; finally, in December, OMB issued a list (#) of Federal policies, regulations, standards, guidelines, and other reference documents pertaining to computer security.</Paragraph> <Paragraph position="10"> The I1baslc philosophy&quot; of the ~omrnunicat ions Act ~ewrite ill remaln the same.&quot; according te former R~D. Louis Frev (R-Fla.). until this vear <. - ranking member of the House ~bmunications .Subcommittee; predictions have also been made that &quot;significant changes: wili be incorporated i-n the legislation this year, previously known as the Communications Act of 1978 (Washington Report, 10J78, p. 3).</Paragraph> <Paragraph position="11"> A new s~ibcommittee on llProfessionalism 6 Malpractice of Computer ,Specialists1I has been formed by the Committee on -Law Relating 'to Cbmputers of the American Bar Association's Science 6 Technology Section; heading the subcommittee is *J.T. Westermeier, Jr., member of a Washington, Q .C. law f inn.</Paragraph> <Paragraph position="12"> FEBRUARY, 1979 8 ~FIPS ViASIIINGTON RrPORT Ed. : Information for the February, 1979, AFIPS Washington Report js current as of January 5, 1979, press time. Production assistance for the Vashington Report is provided by Linda Martin. AFIPS societies have permission to use material in the newsletter for their own publications. Documents indicated by the symbol are available on request to the Washington Office. Requests should specify the da<e(s) of the Report in which the document(s) appeared. Where price is noted, make checks paybble to &quot;AFIPS .I'</Paragraph> </Section> <Section position="22" start_page="81" end_page="81" type="metho"> <SectionTitle> WASHINGTON DEVELOPMENTS PRESIDENT, CONGRESS ADDPESS INFORMATION POLICY ISSUES </SectionTitle> <Paragraph position="0"> Amidst predict ions that- the 96th Congress is concentrating on oversight of existing Government programs, there is no dearth of information policy-related legislation on the Congressional Calendar, sustaining the momentum of the 95th Congress which enacted 74 new laws affecting U. S.</Paragraph> <Paragraph position="1"> informat ion pol icy. [~ditor? s Note : A House of Represent at ives committee Print describing these laws is available on request to the MIPS Washington Off ice. ] Privacy Legislation.</Paragraph> <Paragraph position="2"> Much of the information policy-related legislation ceoters on privacy issues. President Carter referred to planned privacy legislation affecting Government access to records in the medical and financial sectors (see Washington Report, 12/78, p. 1) in his Supplemental State of the Union Address delivered to the Congress on January 25th. Under the heading of &quot;Civil Liberties : Privacy, the President said : Government and private- institutions collect increasingly large amounts of personal data and use them to make many crucial decisions about indfviduals. Much of this 'information is needed to enforce laws, deliver benefits, provide credit, and conduct similar, important services. ~bwever, these interests must be balanced against the individuals right to privacy and against the harm that unfair uses of infarmation can cause. Individuals shoul'd be able to know what information organizations collect and maintain about them; they should be able to correct inaccbrate records; and there should be limits on the disclosure of particularly sensitive personal information.</Paragraph> </Section> class="xml-element"></Paper>