Investigador extranjero visitante:
Tassadit AMGHAR

Fecha de visita:
21/10/07-02/11/07

Actividades a realizar:

Biblio Tassadit Amghar:

I am a lecturer in data processing to the U.F.R Sciences of the University of Angers since September 1998. I carried out my thesis within the LIPN (University Paris 13) and the IRIN (University of Nantes and Ecole Central of Nantes). I occupied a station of ATER at the University of Nantes (1996-1997) then at the University of Angers (1997-1998).

Research topics Tassadit Amghar:

My research tasks relate to the computational linguistics and the knowledge representation. In this context, my step consists in proposing new tools for better taking into account the semantic aspects of the language but also to facilitate the use and the treatment of textual documents within the framework of real applications.

Working project during the visit:

In information retrieval, our work relates to -on the one hand the reformulation and the categorization of the request, and -on the other hand the presentation of the response in the form of composite documents [ACT22, ACT3] as well as the segmentation of the documents given in answers [ACT1, ACT2]. Work on the composition of documents in response to a request with a documentary system involve knowledge concerning document engineering and combinatorial problem solving (optimization, metaheuristics, constraint programming …)
The purpose of this project is to compute as answer not only documents of the corpus but new documents, called composite documents, made up of fragments of the documents provided initially in response to the request. The global evaluation of the quality of the composite documents calls upon criteria of different nature accounting for the synthetic character of the document through the diversity of the information provided in the documents and induce complex combinatorial problems. A first evolutionary algorithm based on a rudimentary segmentation of the documents was carried out and already gave promising results . More recent work concerns the development of text segmentation algorithms using multicriterion optimization in order to determine of the suitable segments.

Datos para contacto en Chile:
e-mail: [email protected]