Decision Procedures For Dependency Parsing Using Graded Constraints

We present an approach to the parsing of dependency structures which brings together the notion of parsing as candidate elimination, the use of graded constraints, and the parallel disambiguation of related structural representations. The approach aims at an increased level of robustness by accepting constraint violations in a controlled way, combining redundant and possibly conflicting information on different representational levels, and facilitating partial parsing as a natural mode of behavior. 1 I n t r o d u c t i o n Language understanding is based on a variety of contributions from different representational levels. From this perspective, one of the most attractive features of dependency based grammar models seems to be their relational nature which allows to accommodate various kinds of relationships in a very similar fashion. Since the basic representational framework is a rather general one it can be (re-)interpreted in many different ways. Thus, dependency relations lend themselves to model the surface syntactic structure of an utterance (with labels like subject-of, direct-object-of, determiner-of, etc.), its thematic structure (with labels like agent-of, theme-of, etc.) and even the referential structure (with labels like referential-identity, partof, possessor-of, etc.). This representational similarity obviates the necessity to integrate too many disparate informational contributions into a single tree-like representation. Instead, representational levels can be separated from each other in a clean manner with appropriate mappings being defined to relate the different components to each other. Another less obvious advantage of dependency formalisms is their suitability for the apSyn