Section: New Results
Participants : Maxime Amblard, William Babonnaud, Philippe de Groote, Bruno Guillaume, Guy Perrier, Sylvain Pogodalla, Valentin Richard.
Abstract Categorial Grammars
Although Abstract Categorial Grammars have well established formal properties that make them suitable for language modeling, some missing features hinder their practical use. For instance, in order to have a compact description of grammatical properties such as number agreement between the subject and the verb of a sentence, a very common approach is to have syntactic descriptions augmented with feature value matrices. Having such a mechanism in Abstract Categorial Grammars requires a lot of attention in order to avoid impacting their computational properties (a previous approach using dependent types showed that, if too general, the problem may become intractable ). We have been working on theoretical approaches to this problem from different perspectives: looking for a computationally adequate type extension of the formalisms, and using the composition capabilities of the framework.
We also have been working on a unifying and general framework, provided by a categorical generalization of Abstract Categorial Grammars . The goal is to get a unified approach to several semantic modeling, and to add numerical methods to the formalism.
Syntax-Semantics Interface as Graph Rewriting
In their book (English version:  and French version: ), Guillaume Bonfante (LORIA, Université de Lorraine), Bruno Guillaume and Guy Perrier devote two chapters to the usage of the Graph Rewriting formalism in the modeling of Syntax-Semantics Interface. Chapter 4 presents two existing semantics formalisms and shows how they can be encoded as graphs: Abstract Meaning Representation (AMR)  and Dependency Minimal Recursion Semantics (DMRS) , . Chapter 5 described two Graph Rewriting Systems proposed by the authors to build semantics graphs in these two formalisms from syntactic dependencies.
The lexicon model underlying Montague semantics is an enumerative model that would assign a meaning to each atomic expression. This model does not exhibit any interesting strucuture. In particular, polysemy problems are considered as homonymy phenomena: a word has as many lexical entries as it has senses, and the semantic relations that might exist between the different meanings of a same word are ignored. To overcome these problems, models of generative lexicons have been proposed in the literature. Implementing these generative models in the realm of the typed -calculus necessitates a calculus with notions of subtyping and type coercion. William Babonnaud is currently developing such a calculus.