How Adaptive Aggregate Contextual Recommender Benifits IP-based TV Services
For achieving personalized content offering, IP-based TV services make use of recommendation systems to automatically suggest programs for users. In such recommenders, to avoid filter bubbles, the research focus is recently transferred to diversity and novelty after many years of pursuing accuracy and personalization. Despite that many approaches have been proved effective to increase the diversity and novelty in recommenders, their defects as unexplainable results and non-interactivity make them not so friendly to users. On the other hand, contextual factors such as timing, location, company by other people, etc., which possess quite clear context meanings are proved influential on IP-based TV services' users' choices. Such contextual factors were often integrated in recommenders though, to my best knowledge, their roles of reasonably increasing diversity and novelty haven't been detailed studied yet. In this Ph.D. work, I plan to realize an Adaptive Aggregate Contextual Recommender for IP-based TV services, which makes uses of multiple contextual factors in an adaptive way to resolve the issues of interpretability and interactivity when increasing diversity and novelty in recommenders.