The rapid growth and diversities of Web information bring a lot of difficulties to efficient information retrieval.The current information retrieval tools just offer-based searching,but ignore the semantic content of the keyword itself.The authors' library information retrieval system takes advantage of ontology,which expands the requirements of users to a semantic word set,and provides a document analyzer to filter the Web pages returned by the search agent according to a certain algorithm.Consequently it presents the most relavant documents to the users.