6533b827fe1ef96bd1285905

RESEARCH PRODUCT

L’ambiguïté anaphorique et la résolution automatique de l’anaphore pronominale

subject

description

Pronominal anaphores can create ambiguities when requesting the correct antecedent. In a context of automation of linguistic and textual processing, several researchers in the field of automatic language processing (TAL) have put in place various approaches to solve the problem. We propose to adapt these approaches to the automatic resolution of pronominal anaphors in a closed body of abstracts of French literary works (RESUMAN).Firstly, linguistic, textual and cognitive models are presented, which report on the functioning and interpretation of personal pronouns in French texts. Then, the interpretation procedures of the pronominalanaphors are exposed : the procedure of minimal distance, the procedure of the parallel functions, the procedure of the subject or the thematic procedure, as well as the procedures of morphological, semantic and pragmatic analysis. Finally, a new version of algorithm, based on a statistical approach, is presented. The RESUMAN tool makes it possible to improve the performance of a TAL in the case of texts that are pronounced pronominal anaphors like those of the corpus-sample studied. The performance of this tool is evaluated and its limits are commented.