Lexical Object Theory: Specification Level

Lexical Objec Theo y: Speci ica ion Le el JOSÉ F. QUESADA CICA (Cen o In o má ico Cien í ico de Andalucía), Se ille, Spain. E-mail: [email p o ec ed] Abs ac . Uni ica ion has become a majo pa adigm in Ma hema ical and Compu a ional Linguis- ics. The esea ch done in his a ea may be classi ied in ou main s eams: ea u e s uc u es as an adequa e model o he desc ip ion o linguis ic phenomena, yped uni ica ion, ep esen a ion o ea u e s uc u es, and uni ica ion algo i hms. This wo k p oposes a new app oach o uni ica ion- based Ma hema ical and Compu a ional Linguis ics: he Lexical Objec Theo y. The main design c i e ia a e based on linguis ic mo i a ion, compu a ional e iciency and o mal soundness. The i s pa o he wo k ou lines he main cha ac e is ics o he Lexical Objec Theo y, i s comp ehensi e o ien a ion, and i s laye ed s uc u e based on he sepa a ion o he ollowing le els: speci ica ion, ans o ma ion, ypi ica ion, ep esen a ion and uni ica ion. The second pa concen a es on he spec- i ica ion le el o he Lexical Objec Theo y. The linguis ic mo i a ion o his model is p esen ed, as well as a de ailed desc ip ion o he speci ica ion o malism, he compu a ional model i is based on, and inally, he in e ence ules on lexical objec s a he speci ica ion le el. Key wo ds: compu a ional ep esen a ion and in e ence, lexical objec heo y, lexical objec speci i- ca ion o malism, uni ica ion-based g amma o malisms 1. Uni ica ion-O ien ed Ma hema ical and Compu a ional Linguis ics Du ing he las ew decades, uni ica ion has become a majo pa adigm (Kuhn, 1962) in Ma hema ical and Compu a ional Linguis ics. Uni ica ion-based g amma o malisms and heo ies sha e a common philosophy o wo k, a ke nel o o mal de ini ions, se e al laye s o linguis ic, ma hema ical and compu a ional heo ies and echniques, and inally, a su ace o ac s ( o mal consequences, applied lin- guis ics, compu a ional esul s, e c.) ha he pa adigm has o ake in o accoun by making use o he c i e ia o cohe ence, eliabili y and e iciency. Since Joan B esnan’s wo k on lexically o ien ed non- ans o ma ional linguis- ics, Ronald Kaplan’s Augmen ed T ansi ion Ne wo ks, and Ma in Kay’s Func- ional G amma onwa ds, he e has been an inc easing use o he uni ica ion p o- cedu e (as a basic ope a ion o manipula ing linguis ic in o ma ion) and com- plex ea u e s uc u es (as a o mal basis o speci ying and ep esen ing linguis ic in o ma ion). The use o ea u e s uc u es has o en appea ed along he his o y o g amma heo ies and o malisms: LFG (B esnan, 1982), GPSG (Gazda e al., 1985), DCG (Pe ei a and Wa en, 1980), FUG (Kay, 1985), PATR–II (Shiebe , 1984), HPSG (Polla d and Sag, 1994), e c. g am1. ex; 22/06/1998; 12:20; p.1 ICPC: PIPS No.: 151505 (g am:g amkap.cls) .1.15 A mo e de ailed analysis o he ield will allow us o classi y he wo k done wi hin i in o ou majo s eams: Fea u e s uc u es as an adequa e model o he desc ip ion o linguis ic phe- nomena: The in e es o ea u e s uc u es as an adequa e model o he desc ip- ion o linguis ic phenomena has mo i a ed he s udy o hei exp essi e powe . As a esul o his, se e al au ho s ha e p oposed di e en ex ensions o he basic model, such as: empla es and lexical ules (Shiebe , 1984), nega ion and disjunc- ion (Ka unen, 1984), non-local alues and co e e ence (Kaspe and Rounds, 1986), e c. In he same way, o he wo ks apply his heo e ical amewo k in o de o ind a solu ion o pa icula linguis ic p oblems, among which we ind sub-ca ego iza ion (Shiebe , 1986), non-local dependencies analysis (Kaplan and Zaenen, 1988) o coo dina ion (Kaplan and Maxwell, 1988). F om a o mal poin o iew, ea u e s uc u es ha e some o he ollowing cha - ac e is ics in common wi h objec -o ien ed o ame-based knowledge ep esen a- ion languages: exp essi e powe , lexibili y, ep esen a ional adequacy and ecu - si i y. In he same way, ea u e s uc u es sha e p ope ies such as ha ing a high le el o abs ac ion, allowing o he de ini ion o classes o ypes, and inco po- a ing mul iple inhe i ance mechanisms wi h objec -o ien ed o malisms (Smolka Aï -Kaci, 1990). These p ope ies show he adequacy o he ea u e s uc u e-based model o he desc ip ion o lexical and syn ac ic hie a chies. The exp essi e capaci y o ea u e s uc u es is equal o ha shown by i s o de e ms in P olog (Mellish, 1992). Howe e , he lexibili y and ep esen a- ional adequacy o ea u e s uc u es p o e hem o be mo e adequa e o he de- sc ip ion o linguis ic phenomena. Fo his eason, some compu a ional ools ha e been designed o pe mi he ansla ion o ea u e s uc u es in o P olog e ms (Schö e , 1993). Fo example: CLE (Alshawi, 1991), ALEP (Alshawi e al., 1991) and P oFIT (E bach, 1995). Typed uni ica ion: Mo e ecen ly, some o malisms based on ea u e s uc u es ha e inco po a ed ypi ica ion mechanisms, ob aining he so-called so ed ( yped) ea u e s uc u es (Ca pen e , 1992; Messege e al., 1990). Some o hese o - malisms a e: TDL (K iege and Schä e , 1994), ALE (Ca pen e and Penn, 1994), CUF (Dö e and Do na, 1993) and TFS (Emele and Zajac, 1990; Emele, 1994). In his con ex , and wi hin logic p og amming, some languages (Li e (Aï -Kaci and Lincoln, 1989) and Oz (Smolka e al., 1995)) use yped e ms. Rep esen a ion o ea u e s uc u es: The in e es o ea u e s uc u es goes be- yond he me e desc ip ion o linguis ic phenomena. O he wo ks s udy he ep e- sen a ion o ea u e s uc u es. Speci ically, hey s udy he deno a ional seman ics o he model: Shiebe , (1984); Pe ei a and Shiebe , (1984); he compu a ionally mos e icien da a s uc u es which pe mi he ep esen a ion o his in o ma ion: g am1. ex; 22/06/1998; 12:20; p.2