PhD defence Meghdad Farahmand
Le jury de thèse est composé de:
- Prof. Stéphane MARCHAND-MAILLET, Directeur de thèse, Faculté des sciences, Université de Genève
- Prof. Gilles FALQUET, co-directeur de thèse, GSEM, Université de Genève
- Prof. Paola MERLO, Faculté des lettres, Université de Genève
- Dr. Andrei POPESCU-BELIS, Idiap Research Institute/EPFL
Résumé:
Idiomatic or multiword expressions (MWEs) are sequences of words (sometimes with a gap) that can be semantically, syntactically, or statistically idiomatic (idiosyncratic). Examples of MWEs are flea market and kangaroo court with semantic idiosyncrasy, at large and in short with syntactic idiosyncrasy and drug dealer and finish line with statistical idiosyncrasy. MWEs constitute an interesting topic in natural language processing (NLP) because while idiosyncrasy cannot be readily defined, it still excessively comes up in any language data, whether it is written or spoken and poses major challenges to many NLP applications. Researchers believe that MWEs account for a significant fraction of daily interactions. According to some researchers, MWEs are as many as single words in a speaker's lexicon, but, others believe that even this is an underestimate. In my work, I discuss idiosyncrasy from a computational perspective. I study idiosyncrasy from a statistical and probabilistic viewpoint and explore and adapt different machine learning algorithms for learning this phenomenon.
Date: Vendredi 17 mars 2017 à 16h00
Lieu: Battelle bâtiment A - Salle de cours 404-407 (3ème étage)
March 17, 2017News