Bienvenue Welcome

Eric Villemonte de la Clergerie
Researcher, member of new Team Alpage at INRIA,
Former Scientific leader of Project-team Atoll

Phone: +33 1 39 63 54 10 (soon January 2016: +33 1 80 49 42 68)
Fax: +33 1 39 63 53 30

Warning As you may see, this personal page is largely outdated (more or less since 2007) ! But promised, someday I will find the time and the energy to update it ... but in meantime please check on FRMG Wiki.

My main area of research deals with the use of tabulation techniques in Computational Linguistics and Logic Programming, relying on the notions of Push-Down Automata and Dynamic Programming. I wish to show that many linguistic formalisms may benefit from well defined and understood tabulation mechanisms, for instance in the case of Tree Adjoining Grammars (TAG), and, more generally, of Midly Context Sensitive (MCS) formalisms using Thread Automata. This work is completed by the development of system DyALog.

Since 2003, I became more and more interested by the notion of Meta-Grammar that provides an abstract layer of syntactic description based on hierachies of classes grouping elementary constraints and providing/requiring resources. These classes may be ``compiled'' to get a grammar, in some target formalism, such as TAGs or LFGs. This interest has led to the development of a MG compiler MGCOMP and of FRMG, a French MG that uses factorization operators provided by DyALog to produce a very compact TAG. New More information about FRMG may be found on FRMG Wiki.

I am also interested in developing infrastructures to help natural language processing (NLP). One way is to use nice representation formats based on XML to encode linguistics resources (grammars, meta-grammars, shared derivation forests, shared dependency forests, morpho-syntactic annotations). In relation with this line of work, I became involved in an international effort driven by ISO TC37 SC4 to standardize linguistic resources.

I am also involved in information extraction and knowledge acquisition applications, starting from parsing output. These thematics include the exploration of error mining and correction techniques and the processing of botanical corpora.