Personalized Video Recommendation Through Graph Propagation

With the development of digital equipment and the advent of near-ubiquitous broadbandInternet connection, users can easily shoot videos using cell phones or cameras and uploadthem to the internet. The video contents on the Internet are growing at a rapid rate. Thesevideos can provide enormous potential for users to find content of interest.However, the vast quantity of videos also turns the finding process into a difficult task.Search is a commonly used technique to overcome the information overload problem. Itworks satisfactory if a user formulates a good textual query. However, in practice it is notalways the case as it is uneasy to describe the search intent using a few textual words. Inaddition, traditional search provides all the users with the same results, ignoring thepreferences of each individual user.To address the information overload problem, we propose a personalized videorecommendation algorithm in this paper. Rather than only exploring the user-video bipartitegraph that is formulated using click information, we first combine the clicks and queriesinformation to build a tripartite graph. In the tripartite graph, the query nodes act as bridges toconnect user nodes and video nodes. Then, to further enrich the connections between usersand videos, three sub-graphs between the same kinds of nodes are added to the tripartite graphby exploring content-based information (video tags and textual queries). We propose aniterative propagation algorithm over the enhanced graph to compute the preferenceinformation of each user.Experiments conducted on a dataset with1,369users,8,765queries, and17,712videoscollected from a commercial video search engine demonstrate the effectiveness of theproposed method.