Dependency constraints for lexical disambiguation

Abstract

We propose a generic method to perform lexical disambiguation in lexicalized grammatical formalisms. It relies on dependency constraints between words. The soundness of the method is due to invariant properties of the parsing in a given grammar that can be computed statically from the grammar.

Lexical Disambiguation in LTAG Using Left Context

“…As advocated by Bonfante et al (2009), using information about context (e.g., dependency constraints) for improving lexical disambiguation may be applied to several types of lexicalized grammar through the concept of companionship. Similarly, it would be interesting to apply left context abstraction to other formalisms.…”

Lexical Disambiguation in LTAG Using Left Context

“…We now show that the optimization we proposed for LTAG is a special case of the more general optimization framework developed in [2,11].…”

“…In this paper, we adapt an optimization technique proposed by Bonfante et al [3] for interaction grammars [13] to prune the initial search space to LTAG and we illustrate the impact of this optimization on two benchmarks. We also relate our approach to a more general optimization technique proposed by Bonfante et al [2] and Morey [11].…”

“…In Section 4 and 5, we present the optimization proposed and we illustrate its impact on two small benchmarks. In Section 6, we relate our approach to the more general approach of companions introduced in [2,11]. Section 7 concludes.…”

Frigram: a French Interaction Grammar

“…This mechanism uses the Companion Property. Originally, this property was introduced by Bonfante et al (2009) A PTD is called consistent when it does not contain any unsatisfiable constraint; in other words, there exists some context in which the PTD can produce a model.…”