Adaptive program filtering under vector space model and relevance feedback

The overabundance of DTV (digital television) programs precipitates a need for smart "filters" to help people obtain programs that they really like. In this paper, we propose an adaptive program filtering system, which is designed to assist users by adapting to their personal preferences. We firstly provide architecture of the program filtering system. Secondly, we present the user profile and program feature representation model and similarity measurement using vector space model. Thirdly, we describe the user profile learning algorithm based on relevance feedback. For user profile learning, we first put forward a primary learning algorithm. With several issues in further consideration, we then present the improved learning algorithm, which is more reasonable and comprehensive than the primary one. Finally, we present the performance evaluation on the prototype of the system.