Section: New Results

Towards a better understanding of frequency effects in syntax

Participants : Benoit Crabbé, Juliette Thuilier.

Some members of Alpage are involved in the statistical parsing of French, the idea of using probabilistic devices for parsing is rather new in France. Alpage has shown earlier [79] that such parsers are performing quite well on French.

Since earlier non statistical parsers were inspired by a trend in linguistics that rejects the idea of granting any importance to frequency effects, it remains largely unknown which are these probabilistic factors that help parsing. This question has almost never been adressed for French. We decided to launch a theoretical investigation aimed at identifying which factors come into play when we take frequency effects into account.

In collaboration with Gwen Fox (Université Pars 3), the first investigation in this direction has been led towards identifying the importance of constraints that drive the placement of adjectives wrt the noun in the noun phrase in French. This study brings an additional element to Bresnan's thesis [70] , according to which the syntactic competence of human beings is indeed probabilistic. Further ongoing studies on adjectives will try to bring evidences for the facts (1) that the grammar of a natural language is intrisically redundant and (2) that we indeed store in our mind not only words of the language but also highly frequent grammatically compositional sequences.

As can be seen from the outline above, this line of research brings us closer to cognitive sciences and more specifically to frameworks inspired by construction grammar. We hope in the very long run that these investigations will bring further insights on the design of probabilistic parsers. In NLP the framework that is closest to implementing construction grammar is Data Oriented Parsing [62] .