Abstract: At present, the similarity calculation for inquires is usually considered by comparing retrieval results to inquires. This paper proposes an algorithm based on Bayesian classifier to calculate the similarity of XML search results. On the basis of working out similarity of each document and inquire, it divides XML retrieval documents into relevant sets and uncorrelated sets by using Bayesian classifier. Then, final similarity is obtained by calculating the similarity of relevant documents and uncorrelated documents. At last, the experimental analysis shows that the new algorithm improves the retrieval performance effectively about 15 percent higher than traditional method without affecting recall ratio.

Key words: Bayesian classifier, inquire similarity, XML retrieval document, information retrieval