The Nature and Structure of a Computational Linguistic Theory

Within the past decade, grammar formalisms based upon the unification of feature structures have come to play a central role in many different research traditions in theoretical and computational linguistics.1 Within linguistic theory, perhaps the most familiar examples come from generalized phrase structure grammar (GPSG, Gazdar et al., 1985), where feature structures are employed to model syntactic categories, and from lexical-functional grammar (LFG, Bresnan, ed., 1982), where they 1nodel a posited level of syntactic representation called (-structure (functional structure) which contains information about the grammatical relations of an expression. Somewhat simplified examples of such feature structures are indicated by the attribute-value matrix (AVM) diagrams in Figs. 1 and 2: Figure 1: GPSG category for the pronoun 'she' Figure 2: LFG f-structure for 'John tried to leave' But it is important to be aware that feature structures can be used to model a very wide variety of linguistic objects (not just syntactic ones), depending on the subject matter of the theory involved.