The Research Of Personal Information Search Technology

In recent years, high-speed expansion of network makes people convenient on getting information, but because the amount of information network is too large, and traditional information retrieval techniques can't meet people's needs, so people can't get information that they want to get quickly. It provides a new challenge on information services. Researchers should consider the user's personality, and try to improve on traditional technology, users'interests are the starting point for personalized information retrieval, interests and background of user are considered. Personalized information retrieval can provide different results depending on different user.After the study and analysis of the related technologies in personal modeling, this thesis provides a detail design of the client-based model. Without much participation of the users, though mining the web pages visited by the users, the users'interests are automatically deduced from the user's implicit feedback.Some technologies on interests mining are studied in this thesis, the model gets users' interests from the pages that they have visited before. In this process, people extract the contents of pages according to structural characteristics of HTML, and use methods based on string matching and statistics law to segment the words in pages, then delete the stop words. In this thesis, page themes are expressed with VSM(Vector Space Model), weight of feature words need some factors to computer, for example, word frequency and location, and used a nonlinear function.A classification model is established, with consulting the ODP catalog structure, and adds feature words, then determine the page's class according to similarity of page and classification model.In the process of updating interests'model, personal model will rectify interest class and feature words, and rectify the weight of them, and used a time forgetting mechanism.When user searches information, search term will be expanded according to interests' model, then use search engine to get the information, the result is better.