Research On Personalized TV Programs Search

Watching TV programs is the most popular way of spending our free time in China. But with the high development of TV technology, there're so many kinds of TV programs around us that it is really difficult for us to choose the programs we like in a short time. Since most TV programs have subtitles, the search of TV programs can be converted to do search on the subtitles. In order to help people solve the problem, a joint research whose aim is to help users find their favorite TV programs rapidly is carried by Ubiquitous Digital lab of Shanghai Jiao Tong University and Hitachi. A small-personalized search system prototype for TV Programs is developed to demonstrate one of the applications of the technique presented in this paper. In the end of this paper, we show that such personalization algorithms can significantly improve search results.Nowadays search engines are widely used by people all over the word. While current web search engines do a good job in retrieving results to satisfy certain people's needs for a given query, they do not do a very good job in discerning individuals'search goals. As a result, users can get thousands of search results, but few results meet user's need. The reason is that keywords that we use to search are not always an appropriate means of locating the information in which a user is interested. Assuming that the system knows sufficient information about user's personal preference, it goes with saying that it can provide them much better search results. Since different people have different interpretations of what is relevant, personalized search is trying to show different results that are most likely relevant to different people on the same search.In order to provide personalize service for users, we use a compound methods of user profile, user feedback and query expansion. We mainly employ user profile to do the personalized search. The system maintains the profiles of its users that represent current user's multiple interests. User's interests which are learned through positive and negative user feedback. According to these profiles, a set of relevant documents retrieved from the document repository is recommended to it users. Moreover, our research suggests that query expansion is also of great help to the personalized search because we can approximate user's query representations through query expansion.To summarize, my main research work in this thesis reads as follows:Design personalized search architecture for TV programs that uses a compound methods of user profile, user feedback and query expansion.Design the use profile structure which can represent multiple interest categories for a single use and develop a dynamic learning algorithm to adapt to the changing user's interests.Design two schemes to perform query expansion through Corpus and IDF analysis of user query similar log words.Design a high efficiency of index compression algorithm.