Parsing Languages with a Configurator.
Mathieu Estratat, Laurent Hénocque · 2004
Abstract. Recent evolution of linguistic theories heavily rely upon the concept of constraint. Also, several authors have pointed out the similitude existing between the categories of feature-based theories and the notions of objects or frames. We show that a generalization of constraint programs called configuration programs can be applied to natural language parsing. We propose here a systematic transla-tion of the concepts and constraints introduced by property grammars to configuration problems representing specific target languages. We assess the usefulness of this translation by studying first a recursive (context free) language with semantics, then a natural language sub-set with lexical ambiguities. Our proposal improves over property grammars because the search procedure in our case is generic and does not rely upon an ad-hoc solver. Configuration techniques also extend constraint programming through object orientedness and set variables. Even though dependency grammars exploit set variables themselves, we foresee that the possibility of coupling the parser with constrained object models describing the word semantics is at the ad-vantage of configuration. It also provides natural (if not easy) parser integration of natural language semantics. Our experiments show the practical efficiency of this approach, which does not require the use of ad hoc algorithms and can be freely used in analysis, generative, or hybrid mode. 1