Research And Realization Of Personalized Search Engine

Internet brings convenience to people. However, data in Internet is giant, and is growing rapidly. So, sometimes it is not easy to find the information needed from the data as sea. A search engine appears, which brings convenience to people.The traditional search engine is mature, and gains a great success commercially. However, the traditional search engine is generally based on matching keywords, and little users'personalized information is considered, which is lack of personalized feature, In order to solve the problem, a personalized search engine is presented to make use of the personalized information, which has personalized feature and can meet users'personalized information need.In this thesis, the personalized search engine is based on the traditional search engine, and personalized feature is added to meet users'personalized information need. The personalized search engine uses a vector space model as a user interest model to specify users'interest and behavior. In the personalized search engine, the user interest model is built, updated and used to optimize queries and sort results, which can realize the personalization. The user interest model is the key to realize the personalization for the personalized search engine, and is the key of the research in this thesis as well.The user interest model is presented as a vector which consists of terms and their weights. The terms and their weights can specify users'interest and behavior, which can be got in such ways as some interest keywords can be provided as terms when a user registers, terms can be extracted from the queries input by a user, and terms can be extracted from the web documents browsed by a user in the way of the user click. Get users'interest and behavior, and build the user interest model. In order to specify users'interest and behavior better, the user interest model need be updated. The user interest model can be used to optimize queries and sort results. To optimize queries, for a user, the terms in the user interest model are sorted in descending order according to their weights, some terms with greater weights are listed for the user to select, and the selected terms are used for retrieving with the queries input by the user together. To sort results, for a user, calculate the similarity of every web document and the user interest model, and sort results in descending order according to the similarity.Based on the traditional search engine, a personalized search engine named PSE is realized in Java, by using Lucene and the user interest model. The personalized search engine PSE makes use of users'personalized information, which can meet users'personalized information need by building, updating and using a user interest model to optimize queries and sort results.The research in this thesis shows that users'personalized information need can be met by applying the user interest model in the personalized search engine.