Guiding a Constraint Dependency Parser with Supertags

Proceedings of the 21st International Conference on Computational Linguistics and 44th Annual Meeting of the ACL, pages 289–296, Sydney, July 2006. c⃝2006 Association for Computational Linguistics Guiding a Constraint Dependency Parser with Supertags Kilian Foth, Tomas By, and Wolfgang Menzel Department f¨ur Informatik, Universit¨at Hamburg, Germany foth|by|menzel@informatik.uni-hamburg.de Abstract We investigate the utility of supertag infor- mation for guiding an existing dependency parser of German. Using weighted con- straints to integrate the additionally avail- able information, the decision process of the parser is influenced by changing its preferences, without excluding alternative structural interpretations from being con- sidered. The paper reports on a series of experiments using varying models of su- pertags that significantly increase the pars- ing accuracy. In addition, an upper bound on the accuracy that can be achieved with perfect supertags is estimated. 1 Introduction Supertagging is based on the combination of two powerful and influential ideas of natural language processing: On the one hand, parsing is (at least partially) reduced to a decision on the optimal se- quence of categories, a problem for which efficient and easily trainable procedures exist. On the other hand, supertagging exploits complex categories, i.e. tree fragments which much better reflect the mutual compatibility between neighbouring lexi- cal items than say part-of-speech tags. Bangalore and Joshi (1999) derived the notion of supertag within the framework of Lexicalized Tree-Adjoining Grammars (LTAG) (Schabes and Joshi, 1991). They considered supertagging a pro- cess of almost parsing, since all that needs to be done after having a sufficiently reliable sequence of supertags available is to decide on their combi- nation into a spanning tree for the complete sen- tence. Thus the approach lends itself easily to pre- processing sentences or filtering parsing results with the goal of guiding the parser or reducing its output ambiguity. Nasr and Rambow (2004) estimated that perfect supertag information already provides for a pars- ing accuracy of 98% if a correct supertag assign- ment were available. Unfortunately, perfectly re- liable supertag information cannot be expected; usually this uncertainty is compensated by run- ning the tagger in multi-tagging mode, expecting that the reliability can be increased by not forcing the tagger to take unreliable decisions but instead offering a set of alternatives from which a subse- quent processing component can choose. A grammar formalism which seems particularly well suited to decompose structural descriptions into lexicalized tree fragments is dependency grammar. It allows us to define supertags on differ- ent levels of granularity (White, 2000; Wang and Harper, 2002), thus facilitating a fine grained anal- ysis of how the different aspects of supertag in- formation influence the parsing behaviour. In the following we will use this characteristic to study in more detail the utility of different kinds of su- pertag information for guiding the parsing process. Usually supertags are combined with a parser in a filtering mode, i.e. parsing hypotheses which are not compatible with the supertag predic- tions are simply discarded. Drawing on the abil- ity of Weighted Constraint Dependency Grammar (WCDG) (Schr¨oder et …

Content truncated to 500 words