Comparable corpora, that gather texts in different languages that are not translations, are multilingual textual resources more and more available on the web.

Metricc proposes to exploit the possibilities offered by these corpora in three industrial applications: translation memory, cross-language information retrieval and multilingual categorization. This project addresses the issue of comparable corpora in a complete and original way. It aims to answer several fundamental challenges for the construction of comparable corpora, the extraction of bilingual resources and their exploitation in the previous applications.