Traitement des erreurs d'accord. Une analyse syntagmatique pour la détection et une analyse multicritère pour la correction

In this paper, we propose an original method for agreement error detection and correction that we apply to Arabic. The detection is based on a global syntactic analysis of the sentence. It consists of an extended syntagmatic analysis that can first locate the syntagm boundaries and, second, regroup all the sentence components that have an agreement relationship. The correction is based on a multicriteria analysis aiming to rank the correction alternatives in order to choose the best one. This method has the advantage of reducing the dominated alternatives and ranking the remaining ones according to different evaluation criteria.