TV Content Search And Recommendation For The Integration Of Three Networks

With continuous integration of the Internet, broadcasting network and telecommunication network, TV viewers can watch on-demand videos and live TV shows more and more convenient. Though the integration of three networks enables the viewer to access lots of TV programs, people tend not to know which video or TV channel they should select facing with a huge variety of TV contents. Meanwhile, people still have to spend much time and energy to find their interested videos from the results of video search engine. Therefore, it is significant to mine TV viewers’ preferences and to provide them with personalized video search and recommend service.To do research in a real environment of the integration of three networks, the thesis builds a live TV module with peer-to-peer function for findball website. According to the request of describing video content, the thesis uses three ways to enrich video tag-based description file. On the basis of a systemantic analysis of current researches on personalized service, the paper proposes a new user preference mining algorithm which has synthesized user’s explicit and implicit feedback information. After examing the limitations of previous cosine similarity-based personalized service, a novel method is proposed to facilitate the desired personalization in video search and recommendation. Results of the research are applied to findball website to provide users with personalized video search and recommend service.The experiments show that the proposed method for personalized video search and recommend service gets better results than other existing methods. However, the personalized video search only takes search results reranking into consideration and video recommend service has limitation of novelty, future research will focus on the two directions.