Ambiguity rates
Résumé
We analysed a French textual corpus in order to evaluate its rate of lexical ambiguity (number of lexical tags per word). Since this rate theoretically depends on the tagset and on whether compounds are delimited by tagging, the experiment was repeated with eight different tagsets. The results show that, although the information content of the tags is very different depending on the tagsets, the variation of the rate of lexical ambiguity is limited: when one shifts from the least to the most informative of the tagsets, the rate increases only from 1.6 to 2.0 tags per word.
Origine | Fichiers produits par l'(les) auteur(s) |
---|