Coping With Dynamic Syntactic Strategies: An Experimental Environment For An Experimental Parser

An envi ronment bui l t a round WEDNESDAY 2, a char t based parse r is in t roduced . The e n v i r o n m e n t is in par t icular oriented towards exploring dynamic aspects of parsing. It includses a number of specialized tools t ha t consent an easy, graphics-based interact ion with the parser. It is shown in par t icular how a combinat ion of the characteris t ics of the parser (based on the lexicon and on dynamic unification) and of the e n v i r o n m e n t allow a nonspecialized user to explore heurist ics t h a t may al ter the basica control of the system. In this way, for instance, a psychol ingu is t m ay explore ideas on human parsing strategies, or a "language engineer" may find useful heuristics for parsing within a particular application. 1. I n t r o d u c t i o n Computer-based e n v i r o n m e n t s for the l i ngu i s t are conceived as sophist icated workbenches , b u i l t on AI works t a t i ons a r o u n d a specif ic p a r s e r , w h e r e the l inguist can t ry out his /her ideas about a g r a m m a r for a cer ta in na tura l language. In doing so, he/she can take advan tage of rich and easy-to-use g raph ic in t e r faces t h a t "know" about linguistics. Of course behind all th is lies the idea tha t cooperat ion with l inguists will provide be t te r results in NLP. To subs tan t i a te this assumption i t may be recalled t ha t some of the most in teres t ing recent ideas on syntax have been developed by means of jo int c o n t r i b u t i o n s f rom l i n g u i s t s a n d c o m p u t a t i o n a l l inguists . Lex ica l -Func t iona l G r a m m a r [ B r e s n a n & K a p l a n 1982], G P S G [ G a z d a r 1981], F u n c t i o n a l G r a m m a r [Kay 1979], DCG [Pereira & Warren 1980], TAG [Joshi & Levy 1982] are some of these ideas. Instances of the tools in t roduced above a re the LFG environment , which was probably the first of its kind, an e n v i r o n m e n t b u i l t by Ron K a p l a n for L e x i c a l F u n c t i o n a l G r a m m a r s , D P A T R , b u i l t by L a u r i Kar t tunen and conceived as an envi ronment tha t would su i t l ingu i s t s of a n u m b e r of d i f fe ren t schools a l l commit ted to a view of parsing as a" process t ha t makes use of an unification algori thm. We have developed an envi ronment with a somewhat different purpose. Besides a number of tools for en ter ing d a t a in g r a p h i c mode a n d i n s p e c t i n g r e s u l t i n g structures, it provides a means for exper iment ing with s t ra tegies in the course of the parsing process. We th ink t h a t this can be a valuable tool for ga in ing insight in the cognitive aspects of language processing as well as for tai lor ing the behaviour of the processor when used with a par t icular (sub)language. In this way an a t t empt can be made to answer basic questions when following a nondeterminis t ic approach: what heurist ics to apply when facing a cer tain choice point, wha t to do when facing a failure point, i.e. which of the pending processes to act ivate, t ak ing account of information resul t ing from the failure? Of course th is kind of env i ronment makes sense only because the parser i t works on has some character is t ics t ha t make it a psychologically in teres t ing realization. 2. M o t i v a t i o n o f the p a r s e r We shal l classify psychologically mot ivated parsers in three main categories. First: those t h a t embody a s t rong claim on the specification of the general control s t ructure of the h u m a n pars ing mechanism. The authors usually consider the level of basic control of the system as the level they are s imula t ing and are not concerned wi th more par t icular heuristics. An instance of this class of parsers is Marcus's parser [Marcus 1979], based on the claim tha t , basically, parsing is a determinis t ic process: only sentences t ha t we perceive as "surpris ing" (the so cal led g a r d e n pa ths) ac tua l ly imply b a c k t r a c k i n g .