Supertagging: an approach to almost parsing

摘要: In this paper, we have proposed novel methods for robust parsing that integrate the flexibility of linguistically motivated lexical descriptions with robustness statistical techniques. Our thesis is computation linguistic structure can be localized if items are associated rich (supertags) impose complex constraints in a local context. The supertags designed such only those elements on which item imposes appear within given supertag. Further, each as many number different syntactic contexts appear. This makes much larger than when less complex, thus increasing ambiguity parser. But resolved by using distributions supertag co-occurrences collected from corpus parses. We explored these ideas context Lexicalized Tree-Adjoining Grammar (LTAG) framework. LTAG combine both phrase information and dependency single representation. Supertag disambiguation results representation effectively parse (an almost parse), parser need "only" individual supertags. method also used to sentence fragments spoken utterances where disambiguated sequence may not into structure.