User Retrieval Information Collecting And Preference Mining In Soccer Video Search Engine

People are submerged in a vast information ocean because of the rapidly growth ofInternet information. Although the Search Engine can provide convenient informationretrieval service, people still have to spend a lot of time and energy to find their interestedinformation from the millions of search results. Therefore, it is significant to mine user’spreference from the search engine’s interactive information and to provide personalizedsearch service.Based on a systematic analysis of current researches on user preference mining, thethesis proposes a user preference mining algorithm which has synthesized users’ explicitand implicit feedback information. According to the real-time request of preferencemining, the algorithm defines session through a semantic analysis of users’ retrievalinformation, and gets implicit feedback information based on session to afford real-timeuser behavior data for preference mining. The preference label and preference actionextracted from the feedback information are described as the LDM which will be used inthe design of the preference model. HWUG preference model is built on the basis offootball knowledge and will also be used in the preference modeling. Because differentpreference action represent the dissimilar degree of user’s favor, thus the preference actionweights are set distinctively from each other. A time attenuation factor is brought into thehistoric preference information based real-time preference mining. The undetectedpreference information has attenuated weights to reflect the change process of userpreference. Results of the preference mining are applied to www.showball.net to providepersonalized video search and recommend service.The experimental results show that the preference mining algorithm which based onHWUG can commendably mine users’ long-term, mid-term and short-term preferenceinformation. Compared with the original research engine, personalized search has a goodeffect on video recommend and improved the user experience a lot. However, the systemhas only considered text retrieval information, whereas the retrieval pictures submitted byusers have to been incorporated. What’s more, the role and the influence of preferencelabel’s modifiers haven’t been taken into account when doing the preference analysis.Future research will focus on the two directions.