Research On Key Techniques Of Information Services In Federated Digital Library

The information services in federated digital library addressed in this paper are associated with the storage, retrieval, discovery, and usage of enormous distributed digital objects on the Internet reasonably and effectively.In this thesis, the infrastructure of information services in federated digital library is presented. It is composed of the set of services that are necessary to provide basic digital library functionality, which are called core services. As an infrastructure of information services in federated digital library, the digital object architecture is described, which provides the mechanism for permanent storage, secure access, management and index to digital objects. Therefore the distributed enormous digital resources in federated digital library can be transformed into knowledge for users.The information retrieval in digital object architecture is a challenging topic, whose trend is semantic index to federated and distributed resources based on AI technology. Some key techniques for the trend are presented in this paper, including search engine selection in federated information retrieval, intelligent information retrieval based on Bayesian network model, and personalized information services.The main contributions are described as follows:(1) The semantic meanings among the terms are mined by means of association rules discovery and four effective methods of computing the conditional probabilities among terms are proposed. The semantic meanings are stored in the Bayesian network model, so the information retrieval system based on Bayesian network model can accurately search information according to the semantic meaning of user's query.(2) A statistical method based on relevance terms to describe resource is put forward, which improves the method based on the hypothesis of independence among terms. Therefore, the accuracy of search engine selection in federated information retrieval is improved and the complexity of computation to select the search engines is decreased. The method is implemented as follows: after the relevance terms are mined by means of association rules discovery, these terms and their frequencies are stored, so the accuracy of search engine selection can be improved because of the relevance relationship among the terms.