Abstract: In order to solve the problem of lower precision with the traditional way of query expansion, a semantic query expansion method based on word sense disambiguation is proposed, which integrate the semantic expansion tree and the co-occurrence of words.For the query topic shift aroused from word sense ambiguous, the query word sense is confirmed by disambiguation based on WordNet knowledge and their domains information, and then the candidate expansion terms are generated through the expansion semantic tree which is generated from hierarchy of WordNet, the final expansion terms are determined according to the Maximum principle of the whole correlation of candidate expansion terms and all query words.User request is fully expressed by these expansion words.Lastly, experimental results indicate that our method has a higher precision while ensuring the recall.