Research On Ontology Based Personalized Web Information Service Technology

With the explosive growth of Internet and the volume of information published on it, more and more Internet systems provide personalized service to retrieve appropriate documents for satisfying user' s need. Howerver, those sysytems do not design a good mechanism to deal with masses of incoming documents. Therefore, people nowadays suffer problems of difficult information management and personalized service requirement since rapid increading Internet documents. So in this thesis, a solution is proposed , which can implement information classification based on ontology by combining the concept hierarchy of ontology with fuzzy neural network and personalized information service by online analysis on the user' s accessing behaviors and online revisement to the user profile. The following features are investigated.The architecture of personalized Internet information service system with the capability of automatic ontology construction is proposed. The system can implement automatic Internet information classification and personalized information service.The method for automatic construction of domain ontology from a set of collected Internet documents is presented. The ICTCLAS(Chinese accidence analysis system)developped by institute of computation technology of china academy of science is used to implement Chinese automatic word segmentation and tagging. The algorithm of data mining is used to extract meaningful features that represent the main characteristics of the retrieved documents. Then, a novel ideal of virtual concept is proposed to organize the extracted features into specific concepts. Finally, the domain concept hierarchy is contructed automatically by employing the concept inheritance method to represent the ontology. This method is suitable to automatic construction of ontology of all the other domains and valuable to applications.The Web document classification system combining the fuzzy neural network with the concept hierarchy of ontology is developped. In theproposed framework, document classification is implemented by first making fuzzy reasoning in the domain ontology. The inference result is then adopted as partial input of a fuzzy neural network for making final classification. Such a scheme can reduce the efforts in deciding classification rules for Web documents, which is fully dependent on manual work.An adaptive user profile is used to provide personalized service. The proposed adaptive profile is also based on a fuzzy neural network with reinforcement learning ability. It receives user' s browing log including browing sequence, browing time and browing frequency, and predefined user preference as inputs for online learning to track the behavior variation of user. Therefore, the proposed framework can dynamically adjust the personalized service to meet the best expectation of each user..