The Research Of Personalized And Social Meta Search Engine Base-on Ontology

By using existing search engines, it is still difficult for people to find usefulinformation in the world of tremendous amount of Internet information. The firstreason is that there is inconsistency between the speciality, limitation, exclusion ofinformation required by users and limitless decentralization on the distribution ofinformation resources. The second reason is that there is not adaptability between thedefects of existing search engines and the requirements of information services.Because of defects of low recall and precision, existing search engines can't satisfyfully people's demand of search. By combining of several member search engines,Meta search engines can cover the defects of existing search engines in some extents.To meet the demand of users' specialized search, personalized search isintroduced into meta search engine to improve search precision by establishing usermodel. Because meta search engine is base on existing search engines, it also existsthe same defects of existing search engine using machine search. In order to solve thisproblem, in the few years, along with the development of Web 2.0 technology, theconcept of social search was put forward. Social search is a kind of search based onperson, its target is to get and improve research results by pooling the wisdom of themasses.In this dissertation, personalized search and social search are introduced intometa search engine to meet the demand of users' specialized search.An integrated of personalized search and social search meta search engine modelbased on ontology was presented. The integrated method of personalized search andsocial search was presented to improve search precision.User model was discussed in focus, and a new type of use model was presented.It has the form of user profile reference model, combining individual model andcommunity model, modeled on users' specialties, and updated by reinforcementlearning.In the research field of personalized search, an algorithm of inquiry intentionparsing was proposed based on user interest model. To improve the selection algorithm of member search engines based on mixture learning, a new selectionalgorithm was improved incorporating user partial model of member search engine, ithas the characteristics of personalized selection and fast selection speed. A weightedposition/globule similarity sort algorithm of search results was proposed whichcombines search result's position information and globule similarity based on userinterest model.In the research field of social search, the application field of a PageRankalgorithm which combines document's content was expanded, and the algorithm'sprinciple was used to propose an algorithm named ScoreRand which can calculate theimportance score of users in the interest circle. And a sort algorithm of collectionsearch results was proposed which incorporates user interest model, user collectionmodel and collection's content.