Adaptation d'une Ressource Prédicative pour l'Extraction d'Information
Résumé
In this article, we present a method aiming at building a resource for an information extraction task, from an already existing French predicative lexical resource. We point out the advantages and drawbacks of two predicative resources we worked with: the LADL tables and Volem. We present the reasons why we finally selected Volem as the most interesting resource for the task. Thereafter, we make a list of the needs an information extraction task implies, and how we include missing information in the resource we selected. We evaluate the resource completed by those missing informations, using it in an information extraction task.
Origine : Fichiers produits par l'(les) auteur(s)
Loading...