Abstract:

Relevance feedback method as a method of query expansion has many algorithm models, such as Rocchio relevance feedback algorithm of the Vector Space Model, BIM (binary independence model) of the Probabilistic Model, the RM (relevance model, RM) of the Language Modeling. In order to further improve the accuracy of query, this paper proposes a reordering algorithm combined with subjective Bayes method and relevance feedback technology. The documents are reranked by the extended terms and feedback documents. The extended terms are returned by RM3 algorithm. It used the Language Modeling as a baseline method, the RM3 method as the method of comparison in the experiments. It shows that our method performs well compars with the Language Modeling and RM3 method. The correct rate is better than the Language Modeling and RM3 method on precision at the first N returned documents .

Key words: subjective Bayes method, relevance feedback, reranking