Une plate-forme générique et ouverte pour le traitement des expressions polylexicales (An Open and Generic Framework for the Acquisition of Multiword Expressions) [in French]

An Open and Generic Framework for the Acquisition of Multiword Expressions In this paper, we present and evaluate an open and flexible methodological framework for the automatic acquisition of multiword expressions (MWEs) from monolingual textual corpora. We start with a pratical motivation followed by a theoretical discussion of the behaviour and of the challenges that MWEs pose for NLP applications. Afterwards, we describe the modules of our framework, the overall pipeline and the design choices of the tool implementing the framework. The evaluation of the framework was performed extrinsically based on an application : computerassisted lexicography. This application can benefit from MWE acquisition because the expressions acquired automatically from corpora can both speed up the creation and improve the quality and the coverage of the lexical resources. The promising results of previous and ongoing experiments encourage further investigation about the optimal way to integrate MWE treatment into NLP applications, and particularly into machine translation systems. MOTS-CLÉS : Expressions polylexicales, extraction lexicale, lexique, mesures d’association, corpus, lexicographie.