BiTermEx Un prototype d'extraction de mots composés à partir de documents comparables via la méthode compositionnelle

Emmanuel Planas (1)

1 LINA - Laboratoire d'Informatique de Nantes Atlantique

We describe BiTermEx, a prototype for extracting multiword terms from comparable corpora, using the compositional method. We focus on morphology-based variations of multiword constituents and their recombinaison. We experimented our approach on scientific and popular science corpora. We record TOP1 precisions of 92% and 97,5% on French to English alignments and 94% on French to Japanese.

Domaines: Informatique et langage [cs.CL]

Emmanuel Planas. BiTermEx Un prototype d'extraction de mots composés à partir de documents comparables via la méthode compositionnelle. Proceedings of the Joint Conference JEP-TALN-RECITAL 2012, volume 2: TALN, Sep 2012, Grenoble, France, France. pp.415-422. ⟨hal-00727813⟩