Abstract:

Mutual Information has been widely applied to many expanded information retrieval models.Aiming at problems in mutual information,for instance,being apt to lowfrequency words and ignoring negative potential impact led by sparse data,this paper applies improved mutual information in an expanded information retrieval model. Experimental results on the five normal datasets show that the expanded information retrieval model based on improved mutual information outperforms that based on traditional mutual information.

Key words: query expansion;mutual information;information retrieval