Exploring synonymy relation between multi-word terms in distributional semantic models
Résumé
Terminology describes the knowledge structure of a domain through the relationships between its terms. However, relations between multi-word terms (MWTs) are often underrepresented in terminology resources. Moreover, most of the work on this issue concerns the relations between simple terms (STs). In this paper, we explore the ability of distributional semantic models (DSMs) to capture synonymy between MWTs by lexical substitution based and analogy based methods. We evaluated our methods on the English and French MWTs of the environmental domain. Our experiments show that the results obtained using analogy in static word embeddings are globally better than the ones obtained using lexical substitution in pre-trained contextual models.
| Origine | Fichiers éditeurs autorisés sur une archive ouverte |
|---|---|
| Licence |
