Comparability Measurement for Terminology Extraction
Résumé
In this paper we describe recent work carried out in the context of the TTC project towards the automatic construction of comparable corpora for multilingual terminology extraction. We focus on the communicative intention as the variable of discourse analysis that is best suited to select Web documents valuable for terminology applications and propose a classifier based on language independent features to automatically cluster crawled documents sharing the same communicative intention. The results of our experiments indicate the need to consider more sophisticated features.
| Origine | Fichiers produits par l'(les) auteur(s) |
|---|
Loading...