Semantic and Contextual Knowledge Representation for Lexical Disambiguation: Case of Arabic-French Query Translation

We present in this paper, an automatic query translation system in cross-language information retrieval (ArabicFrench). For the lexical disambiguation, our system combines between two resources: a bilingual dictionary and a parallel corpus. To select the best translation, our method is based on a correspondence measure between two semantic networks. The first one represents the senses of ambiguous terms of the query. The second one is a semantic network contextually enriched, representing the collection of sentences responding to the query. This collection forms the knowledge base of our disambiguation method and it is obtained by alignment with the relevant sentences in Arabic. The evaluation of the proposed system shows the advantage of the contextual enrichment on the quality of the translation. We obtained a high precision, relatively proportional to the precision provided by the used alignment. Finally, our translation demonstrates its potential by comparing its Bleu score with that of Google translate.