Seminaire: A. Korhonen, le 11 juillet au LIPN

[Thierry Hamon]

Date: Mon, 9 Jul 2007 10:59:03 +0200
From: Thierry Poibeau <Thierry.Poibeau at lipn.univ-paris13.fr>
Message-Id: <1711c5db626f62356b6ac17cf7f0750a at lipn.univ-paris13.fr>
X-url: http://www.cl.cam.ac.uk/~alk23/
X-url: http://www-lipn.univ-paris13.fr/planfac/?lang=fr


Anna Korhonen (http://www.cl.cam.ac.uk/~alk23/), de l'Université de
Cambridge, chercheur invitée au Laboratoire d'Informatique de
Paris-Nord (LIPN), donnera un séminaire sur des techniques de
classifications lexicale pour les applications de TAL. Le séminaire
aura lieu le mercredi 11 juillet 2007,au LIPN, à partir de 15 heures,
en salle B311.

Pour venir au LIPN : http://www-lipn.univ-paris13.fr/planfac/?lang=fr

---

Automatic Lexical Classification for NLP Applications

Words that are similar in terms of their meaning and syntactic
behavior can be grouped into lexical classes (e.g. Levin, 1993). Such
classes capture useful generalizations over a range of
(cross-)linguistic properties. They can therefore be used to support
various (multilingual) NLP application tasks which require/involve
abstracting away from individual words (e.g. language generation,
information extraction, machine translation). For example, lexical
classes can be used to build a lexical organization which effectively
captures generalizations over words and predicts much of the behaviour
of a new word simply by associating it with an appropriate
class. However, to date the use of lexical classes in NLP has been
limited because no technology for comprehensive and accurate
(i.e. automatic) word classification into lexical classes is
available. In this talk, I will describe work which we have done
recently in Cambridge on the development of such technology. I will
first describe our method which involves classifying syntactic
features in corpus data using various unsupervised and supervised
machine learning methods. I will then report experiments focussed on
classifying verbs in both cross-domain and domain-specific
(biomedical) texts and evaluating the classification performance
against manually built gold standards.


-------------------------------------------------------------------------
Message diffuse par la liste Langage Naturel <LN at cines.fr>
Informations, abonnement : http://www.atala.org/article.php3?id_article=48
English version       : 
Archives                 : http://listserv.linguistlist.org/archives/ln.html
                                http://liste.cines.fr/info/ln

La liste LN est parrainee par l'ATALA (Association pour le Traitement
Automatique des Langues)
Information et adhesion  : http://www.atala.org/
-------------------------------------------------------------------------