Nội dung

SOLUTIONS TO ISSUES DEPEND ON THE KNOWLEDGE REPRESENTATION Frederick B. T h o ~ p s o H ~li£orni~ I n s t i t u t e o£ T e c h n o l o g y Pasadena, C~,li?orni~ In o r p Q n i z i n g This p ~ n e l , our Ch(tirmon, Bob Moore, expressed the view thor too often discussion o? H o t u r r a ' l l,',nguage o c c e s s To dol'o b u s e s h a s f o c u s e d on whot p ~ r t i c u l c ~ r systems c~*n o r c Q n n o t do, ro'ther t h a n on u n d e r l y i n g issues. He Then s d ~ i r r 4 b l y proceeded to orgonize the p r J n e l nr. o u n d i s s u e s r-qther th~n s y s t e m s . In r e s p o n d i n g , I q t t e m p t e d to ?rr.iMe my ~'emr~rk~, on e,ach o? his f i v e i s s u e s in r~ g e n e r ~ l woy that w o u l d not r e f l e c ~ ~y ,wn pr4rochiul e x p e r i e n c e qnd i n t e r e s t , At one p o i n t I t h o u g h t th~.~t I h~d s,cceeded q u i t e w e l l . Howe,.,er~ o f f e r t ~ k i n g a c l e o r e r eyed view~ it wqs q p p a r e n t thor my r e m a r k s reflec~c;d ~ s s u m p t i o n s obout knowledge r e p r e s e n t o t i o n t h e f t were b y no M e o n s univers~ol. T h i s s u g g e s t s ,a s i x t h issue which I would l i k e ~o r ~ o m i n ~ t , . , : , ' ~ d d i t i o n , i l item o? i n ~ ' o r m o t i o n , n o m e l y ~ h e i r r~umber, Thus ~ DEPARTMENT c o u l d h~ve ~s o s i n g l e ~ember su,:h ~ c o l l e c t i v e ,.,f employees, indeed i t c o u l d hove severQl such c o l l e c t i v e MeMbers ond o t h e r i n d i u i d u ~ l MeMbers ~s w e l l . The procedure t h o r i s c~11ed when onswering "how Mony" ~nd "nu(~ber o f " q u e s t i o n s would know the d i f ? e r e n ( - e between subcl~s,~es, i n d i u i d u r J l (~eMber~ r~nd c o l l e c t i v e ~ e m b e r s ~ it w o u l d know to recurse on s u b c l c ~ s s e s , ~.~dd o n e t o its coun'~ f o r individual MeMbers ond odd t h e indicoted' number" t o its c o u n t - f o r ' c o l l e c : t J v e MeMbers. This ~ppe~rs To be u n i ? i e d ?r,~ework th,,t w i l l h(1~d].e ,~11 o f t h e c:~s,e~ m e n t i o n e d i n Bob M o o r e ' s stQtement o? Issue # I . Issue Are t h e r e r ' e o l l y useful g e n e r o l i z N t i o n s ~bou~ comput~Jtionr~l l i n g u i s t i c issues th<~t c~r,e i n d e p e n d e r ~ t o f r ~ s s u m p t i o n s concerning knowledq,.~ r ' e p r e . s e n t c i t i o n ? I w i l l come back q?ter discussing Choir~or,. '~2: .T,,iM,e, qnci T_e_p._s_e I should l i k e to s p l i t t h i s issue i n t o two, The ? i r s t s u b - i s s u e i s the problem of hondling continuusly varying phenoMeno> ~.~uch ra,:, 1he M o v e M e n t o f s h i p s ~ the chqnging o f relotiv,., ~zMount,:~ o f i n g r e d i e n t s in c h e ~ i c ~ l r e Q c ~ i o n s ~ or "the percent completions o9 t n s k s , Here it is ~pp(~rent thr~t eoch i n s t o n c e will r e q u i r e s ~ p e c i ~ l i z e d procedure to hKLnflle i n t e r p o l , ~ t i o n . Ships cr~nnot s ~ i l Qcross Irjnd~ t h u s ~n i n t e r p o l a t i o n procedure t h o r produce~ the position o f o s h i p on The bdJsis o f its p o i n t s o~' d e p ~ u ' t u r ~ , o n d des'tinotJ, orl w i l l need To know obout the c0cAstlines o? c o n r i n e n t s ~ M o v e m e n t s to cheMJ.c~l e q u i l i b r i u m s . ~re not l i n e ~ r ~ t~sk coMpletioNs depend on changing p e r s o n n e l ossignf~ents. Just rls we c o M p u t o t i o n o l linguists provide t o o u r syste(,~ u ~ e r t h e (.opobili'¢y to introduce into his dotr~ b~se s y s t e m ~ u c h n o t i o n . ' ~ ,a~ l o c , = r i o n s of" port.~ end s h i p ~ , e t c . > we Must ,also p r o v i d e ~he Met:ins by which he crarl d e f i n e such '.:ontinuously vr~ryir',~ p , a r ( ~ m e t e r s r~s position i n s u c h w q y s th~.~t ~ t p p r o p r J . o t e in~erpo],',tions c~zn be ~ d e . by ~he g e n e r o l system in con.junctioN with the p,lrtlculor defini'tJon. For" example, 'the user mr~y de?ine~ "position o? X " i n t e r ~ of ('.r~lcu]:'~%ions, perhops extensive> involvin 9 ~he ,~ctu~.~l 9 e o ~ e t r ' y o f t h e ~ e q ~ . 'to this s i x t h issue t'i~e ?ire cho-~.,n by our I ~s~,e @i : A.q.ctreq~te_.....F.~nc.t i o ns...qnd ~uon.!i.S.~. F'irst~ l e t us csst t h i s issue in o ~omewhot di.fferent w a y , Irl m~.sny d~tc, b~se ~:ituo.1io~s., there ,',r'e c l o s e s of i n d i v i d u a l s o l I o f w h o s e Me~bers shcire the ~.:oMs o t ' ~ r i b u t e s ~and thus, ÷'rum the p o i n t of view o.f the dr~'t~ bose~ ,ir~ .~.ndi~tJ,vuish,~ble. 'Thus t h e r e i s no need ~o ~dd ,*II o f the~.~., i n d i v i d u a l s Qs ~':.e.prw, o't(~ e n t i t i e s , To use Bob M o o r e ' s ex,a~ple., i? ,:z DEPARTMENT t i l e h ~ ,.~ ~ i e l d Cur NUMPER"OF-EMPL.OYEES~ i t strands To r'~~,.~on th~L~' the pczr!icul~:~r [ndividu~.~Is ;~ho ~ c ~ u o l l y e x i s t e d in the v,ar'ious dep~r't~tHH~'~ w o u l d r~o~ be s~ep,ar,~tely r'~pr'e,>~:~,~ed i n the. dc~~obose (for u t h e r ' w i s e t h e r e would be o redund,.~ncy whose ,..:or1~J'-.:'~.;nc,./would be h~r'd to p o l i c e ) , In ~u,-'h :~i~u,:~io~s we n e e d the. n o t i o n o f ~'~ " , ~ o l l e L ' t i , ~ e , " homely ~a s i n g l e dr~tr~ b,ase object ~'hot ~,.,l(e~ the; pl~.'~c~, of ra n u m b e r o f .ir~divid~.,r, l s end which c,~r~ c , a r ' r y t h e i r cot~on ,~r!'r.Lbul'e~ together with one 169 The second s u b - i s s u e on which I w o u l d t o coMmen~ c o n c e r n s ~hose c ~ s e s where d i s c r e t e ~ime i n t e r v a l s provide rin r.~dequ(l~(~, r, e p r e s e n t r . ~ t i o n o@ "(he t i m e a s p e c t s r e l e v r , n t to t h e d a ' t q b q s e . In *hese coses~ if` ~he t i m e i n f ` o r f ~ t i o n is coMplete~ i,e,, r i c t u ( : l st~Ir~lng ~nd e n d i n g rimes of` rill e v e n t s ripe r e c o r d e d i n ~he d ~ t Q b,~se, the h ~ n d l i n g o? time is rrither strriightforwrird, H o w e v e r this c~se of"~en d o e s not ripply, C o n s i d e r the ? a l l o w i n g e xQmp],e. : Ai: like. dept. One rins~er would in one London be: I n p r e s e n t i n g i~his i s s u e Io the prlnel~ Bob Moore used the ? o l l o w l n g t h r e e q u e s t i o n s ris ~n example: " I s John J o n e s (, c h i l d r l l U MIIU ~- ? oF rln HIT " " I s one o£ John ~ o n e s ~ s p ~ r s n t s on H I T ( l l u~)n u.~ 9" "Did e.i'~her poren~ o~" /ohn ]ones t~ttend HIT?" Choosing ~ relev~mt dic~gnosti,: t,eessoge~ os o b o v e . , i s o M r i j o r r~nd di£?icul't coMputotion,:~l liguis~c issue (:~oing w~],l b e y o u n d q u e s t i o n s concerning t iMe ~lII,J *,ense, The a p p q r e . t problem i s 'the po.~sibilit~,, of" M u l t i p l e des(-rLptions~ o ~ ' t e n i n v o l v i n g d i s p o r ' r i t e words~ .For, g e t t i n g ~lt dril:t, i n ~he datri h~se, In (.JeBicjrlillg our" s y s t e m s ) we r e c o g n i z e two t r u ~ h ~ which ~ppe,~r' t o c o n , f l i c k : ( q ) t h e v~lue o.F M i n i M i z i n g the reduhdrincy o£ .LnforMf~tion in the dqt(t b,1~e. (b) the necessi-iy o£ n o n - i n d e p e n d e n t w o r d s in the vocobulr~ry, In our' own work~ ,~s Mo.~t o? you k n o w , we h o v e s t r e s s e d t h e use o? definitions c~s u Me~zns o f ,'Ich.i.eving o s y n t h e ~ i ~ oF '~-hese *.wo princ:iples. I r e c o M M e . d it to you u~ ri v~.r~ o~p.i~ul tool in h o n d l i n o p r o b l e m s like Bob p r e s e n t s . We illustrate how Bob~s excLMpl~; c,.~n be hr~ndled : _~_~r.ULi.f,,.!..i.]:Lg ~,p_xo...~u.esxi o n s T h i s i~ a deep~ p h i l o s o p h [ c , l l ques'~io,~. CoMput~,i. onol l i n g u i s t s h o v e pr, o g r w : s ~ e d b e y o n d tl'w. , : o n s i d e r q t i o n o? ~:~ing].e ~,.~'ntences, r~nd rire s e e k i n g ~o , ? o l l o w ~che ' f o c u s o f ,] d i r j l o g u ~ .~ (And i d e n ' t i . f y 'the theme o? ~ d i s c o u r s e , This i s ev~.r}~uqlly ,4n i n f i n i t e r e g r ' e , ~ u l t L M r i t e l y i n v o ] . v i ~ g cross c u l ~ ' u r ~ l b r i c k g r o u n d s , ~he ( p e r h ~ / p s Mc,chiclvellion> irlt,~.n~ o~ t h o s e ,.,,ha co~,trol ~he u'~e oT o p a r t i c u l a r ctpplicr:~Lon;. ,:It, Dv~: the eng.i.neering prob].ef~, c~ le,~s'r ~ ~he p r e s e n t eta're o£ •the c~rt,, i~ :~Lapl~_: whr,t response is ~ o s ~ :Jse£u] "~.o ~'he use.r? Consider kwo p o s s i b l e (n~swers ~o ,.'he F o l l o w i n g question: "Who A Issue #4 : ~4e_.,rvi.n~..~.eM~n~ic~.!_l..y_ CoMp ke X "The Ki~i'.vh~.~wk r i r r i v e d IriF~t Mondc~y~ the Mor,u will soil n e x t F r i d a y , I~ they will hrive b~.en there q t the same ~iMe~ t h e n not rill s h i p s ~hr~t were i n London t o g e t h e r w i l l s r i i l 1'ogether~ bu~ t h e y v:ould be the only e x c e p t i o n s , " Issue %3 : ~rin~ger o~ e x p e r i e . , c e , "e~ch" con u s e f u l l y be i n ' t e r p r e t e d ,~s c o i l i n g ~ o r ~, l~b~.led l i s t ,~s onswer i n ~ l ~ o s t oll coses, The diff"icul¢ies of" b e i n g t~ore c l e v e r o r e g r e a t arid will o~°ten r e s u l t in coMbinqtori(ll e x p l o s i o n , I (~M s u r ~ . , for o ]ong tJi~e into the f u t u r e , we will be seeking simple s o l u t i o n s t h a t (?.~) o r e respon,~,ive i n M o s ~ c(~ses> (b) p r o v i d e the needed inforMr~eion~ even though r e d u n d ~ n t Jn SOMe cGse~ rind ( c ) M~ke c : l e r i r the M i s i n t e r p r e ~ r i t i o n in the £ e w c,~se where this rirjses, e v e n t h o u g h these s o l u t i o n s May violrite strict l i n g u i s t i c rinqlysis, One i s teP~pted ~o o l l o w the c o m p u t e r lo give q response: "Possibly," however the i n t r o d u c t i o n of` a three uolued l o g i c i s Tr'qught w i t h w e l l known d~ngers of" i t s "Will s i ~ i p s t h r i t have been i n t o g e t h e r sriil t o g e t h e r ? " A M~nages ~ l l U n l e s s ~ h e r e w e r e ~n undue n u m b e r o~" dep~rtMents involved~ the second is (.Ice, Ply prei~erred~ ?or it ~uf`f`ices "even i? the f ` i r s t were i n t e n d e d , I. our own "The Ki~tyhriwk ~ r r i v e d in L o n d o n Monday, The Mriru will soil ?ram L o n d o n Friday. W i l l th~ K i t c y h o w k ~nd Hrir'u h~ve been in L o n d o n ~t the s ~ e time?" own, A More p r o t r r i c t e d response gets t h e way o~ c l n u s e [ ~ b e d d i n g ; how d o e s hrind].e~ "No s i n g l e person the depor~Ments," "de'fini~ion'child:converse o~" p a r e n t ve.rb:John ",;it~end"~ H I T m J o h . is '.-'tud~.t,~ o? HIT dei'~inition:~lu~,r-'s'person who hod been ~ student" ,.~on,.'~gesec~cl~ d e p o r t m e n t ? " 170 The ubove t h r e e ?.~n~1y zed ~s: questions Issue ~6: t h e n ore "3"ohn ) o n e s i s ( c o n v e r s e o? p a r e n t ) o? a p e r s o n who had been ~ s t u d e n t of H I T ? " "One of ~ohn / o n e s ' s p a r e n t s i s a p e r s o n who had been a s t u d e n t o-t H I T ? " "W~s e i ' t h e r p a r e n t o£ ~ohn ~ones a student of HIT?" As I look back on the abuv~ r e M o r k ~ t:oncerning Sob's five issues~ i t becomes ~ p p a r e n t thr~t t h e u . ~ e f u l n e s s of t h e s e r e m a r k s depends on The degree one i s aware o f t h e knowledge r e p r e s e n t a t L o n that underlie.s the solution suggested, For ex~Mple~ in the case of the l a s t ,Ls~ue,. i l ~ one only knew about t r a d i t i o n a l f i l e s t r u c t u r e s ~ f i n d i n g paths theft l i n k f i e l d s in More Than one f i l e appears a l l b u t unsolvable, Even if one i s a c c u s t o m e d t o s e m a n t i c n e t s t r u c t u r e s ~ the v i a b i l i t y of f i n d i n g connective pnth.~ is h i g h l ~ dependccn~ on t h e e x i s t e n c e of back l i n k s between a t t r i b u t e s and t h e i r ,~rgu~en~s and values. Adding a d e f i n i t i o n a l c a p a b i l i t y ~ o t h e r thun s i m p l e a b b r e v i q t i o n s ~md synonyf4s~ B u r n s on t h e way free v a r i a b l e s ore handled in 9ener~Jl cmd on t h e opporo.tus +'or b i n d i n g theM) f o r example, in processing the d e + ' i n i t i o n : I do not wish to slur' over' the fact that ~.= d e f i n i t i o n Mech~.,nisM ~ust be h i f h l y :sophis~'~coted i n i~s h a n d l i n g of f'ree variables,, bu~ our ~xperience i~dic~*te.~ tha~ ~l'~s can be done q u i t e s ~ t i s f a c t o r i l y. Issue Solution~....tO. Is~q.es_._D~.~.n.. #5: H u ~ t i - F i l # . _ ~ u e r ' i e _ . s This issue has been s t a t e d by Bob in terms of G tr'~dixional M u l t i p l e file de=to b,~se s ' t r u c t u r e , T h i s i s s u e h~s i t s c o u n ' t e r ' p ~ r t in seM~intic neT data. base structures d i s c u s s e d in pr4per,~ on k~ow].edge r e p r e s e n t a t i o n , Since we use such q semantic net s ~ r u c t u r e For, our data, le't me rephrase the issue in those ~erMs. In Dab's st~tteMerlt of the issue~ he uses tl'~ example o f the SHIP f i l e and the PORT .File; wl}ere the SHIP f . i l e h,~s f i e l d s -For ho~,~¢ p o r t , departure p o r t and d e s t i n a t i o n port. P,~,r'allelLnq his exa~p](:, l e t us c o n s i d e r ~h~ phrase: " L o n d o n s h i p " , Suppose ~hr.~t ( q ) there w~s s h i p n,~r~ed L o n d o n , nnd ( b ) L o n d o n was a ho~e por~, port o f d e p q r t u r ~ and des'tir~o'~ion~ not n e c e s s a r i l y o~" t h e same s h i p , Then " L o n d o n s h i p " i s f o u r ways ,~Mbifuous~ ~e~ning: (i) t h e ship L o n d o n ~ (2) London (ho~e port) ships, (3) l.ondon (depr~r~-ur~z par-X) ~hips ~nd (4) London (destinq~ion port) ~hips, In this f o r ~ u l ~ t i o n of t h e p r o b l e M ~ ~II i s easy~ insofar ~ the phr~s~; " L o n d o , ship" is n o t '.iisc~Mbigu~Ted in c o n ' t e x t ~ t h e u s e r i s i n f o r m e d o? t h e ~lMbiguous M ~ l r l i n c j s (Ind t h e ,~ssoci,:~'ted r e s p o n s e s . The d i f f i c u l t y u r i s e s when There ar'~. pos.~ibile .i.nterpr,.'<'~'ations ?,~r~her (~field, F o r t C o l l i n s is n,.~.itl')er ,~ p o r t nor ~ ship~ h o w e v e r ~.he headqunr'ters of t h e ABC S l l i p p i r t , # CoMpany i~ t h e r e un,:l t h e y own ~everol s h i p s . Wh,~'t ?~r'e we ~o ~e~n by "For"t C o l l i n s .~hip"? The.~e ~u-e pr'obleM~ tha't wer.e ? i r s ~ ~1'*:acked by Q u i l l i c m , and f qM n o t ~ur'e ~'t~(~'t u n y o n e I..~.~ c~dded to h i ~ !=emii~ol ~r~r~ly':sj.s o£ l h e ~ , In our own work~ we he*re s~uppecJ a t " o n c e re~-1oved" ,.:onnec.'tJons., ,:~ il].u~r~zTed by t h e f o u r w~,~y ,~mbiq,)ity ,Ibove. " d e i ~ ] n i t i o n : a r e ~ : l e n g t h times w i d t h " when a p p l i e d ~o q class> say " a r e a s of ;~hips", how does one ensure ~hat he will ob rain : " l e n g T h ( i ) ~k w i d t h ( i ) fop i = I to number o f rather ship~" Thonl "lengTh(i) ~ width(j) ?or i~j = i to nut.~ber o? ship..=.?" It c o M e ~ down to h o w v a r i a b l e s qre MainTained in The u n d e r l y i n g knowledge represen'tat ion, O n e i s £or'ced t o c o n c l u d e ~ h a t t h e b a s i s ~'or t h e i n t e g r c l t i o n of the s y n t a x cu,d ~ e m o n T i c s o? coMput,~tionr~l l i n g u i s t i c s y s t e m s i~ - c c o M p l i s h e d wh..n tile d¢ci~4ion~ on k n o w l e d g e r ' e p r e ~ e n ~ t i u n ~r'e Made, D i ~ J c u s s i o n s 0£ #ur w:~rLous sotut.En.n to ~he J~sues of c o M p u t a T i o n a l l i n g u i s t i c s can M e a n i n g f u l l y ~uke pl<~ce only in terM~ uf the,:,~ u n d e r l y i n g k n o w l e d g e repr'eser~tot ions. 171