Pour le (re)démarrage de ce blog, une annonce: Loïc Grobol interviendra tout à l’heure au Inria Junior Seminar pour présenter ses travaux réalisés dans le cadre de sa thèse, co-encadrée par Isabelle Tellier (LATTICE), Marco Dinarelli (LATTICE) et Éric de La Clergerie (ALMAnaCH).
Titre : Deep learning for automatic coreference detection
Résumé : Coreference resolution consists in identifying natural language expressions that refer to the same conceptual entity. During the last years, new methods, relying on deep learning techniques to improve the performance of classical heuristics, have revived the interest for a task that was until then seen as having reached a performance plateau. In this talk, I will describe these methods, and try explain how they can outperform the formerly state-of-the art models, but also why they will not suffice to solve this problem in the general case.