Research On Ontology-based User Model For Personalized Search

People have stepped into information age with the development of Internet. At present the on-line resources grow by the index speed, and the information both in the quantity and in the type are unimaginable. Simultaneously people's need for information is also increasing, and it is difficult for people to obtain what they need from the information sea in an accurate, quick and convenient way. As a result, the problem so-called "information overload" and "information confuse" appears. The general search engine cannot consider the personal factors for every individual. So how to learn about the user intention, obtain the personalized information and provide personalized service for the user have become hot points in the information retrieval field.User model is a key issue for the research on personalized information service. And before we provide personalized services for different users, we must construct user models for them. Most traditional user models roughly divide the user's demand into "interested" and "not interested" according to the user's information needs, and finally they use the keywords to describe the user's interest. However, the traditional methods have not considered the intrinsic relations among words, and can not understand the information semantically and serve for the construction of user interest model with the knowledge system of the information sources.It is thought that, personalized user model should have semantic associations besides the accurate description ability; and it could excavate the user the latent interest from the explicit information semantics.As a concept modeling tool which can describe the system in the semantics and knowledge level, ontology has aroused the domestic and foreign researcher's interest since it was proposed. Being a concept model of domain, the research indicates that ontology can explicitly describe concepts, concept meanings and relations between the concepts, and endow with the explicit background knowledge for the simple terminology. Therefore it not only is of benefit to knowledge sharing and reuse, but also is useful for the semantic expansion and reasoning. It will enhance the relevance of the keywords in the keywords vector, if ontology is introduced into the user model field. Constructing user model based ontology has been a new direction of the user modeling for the personalized information service system.There are some user models which are based on ontology. And they have improved the insufficient semantics flaw of traditional user models, but there also are some problems: (1) the problem of developing domain ontology. How to develop domain ontology is the foundation of user model based ontology. With the traditional method domain ontology is always developed manually and totally relied on users and domain experts. As a result, it easily produces some flaws such as deviation, subjectivism and the empiricism. (2) the problem of updating user model. The existing user model mostly tracks the browsing or retrieval process, extracts and analyses the keywords which are thought of user's interest, then takes these words into the initial user model for supplements, while missing considering the semantics of the new increases.The article proposes a new improved user model-OBUM based on domain ontology. While introducing ontology into user model, it is considered to construct domain ontology by text mining and update OBUM by ontology learning. The research work of this article has mainly included such several aspects:First, the article studies and analyses the characteristics of traditional user model, and proposes a new user model-OBUM based on domain ontology. Then we analyze and design the articheture, and make formalization for OBUM.Second, the article researches on the stage of constructing the OBUM model as well as every parts of OBUM. Including:①how to construct ontology effectively is studied. It is proposed to construct domain ontology DomO semiautomatically by text mining with the help of domain experts. The construction of DomO is implemented with the tools of text mining and ontology constructing.②how to obtain the personal ontology of the OBUM model is studied. It is proposed to obtain the initial personal ontology by ontology projecting on domain ontology and its algorithm description is listed.③in view of user model OBUM updating, it is proposed to implement it by ontology learning. We construct reference ontology by analyzing the browsing or retrieval process, and then merge it into personal ontology to complete the personal ontology learning, which avoids the semantics loss of the new increases in the process of user model updating.In the end, we apply the OBUM model to the personalized information search system based on above work, in order to prove the feasibility and validity of OBUM. The experiment result indicates that OBUM model is an effective and feasible method in enhancing user personal information and improving the semantics, and the recall and precision ratio are improved compared with traditional user model based statistics and existing general user model based on ontology in the information retrieval system.