Twitter User Modeling and Tweets Recommendation Based on Wikipedia Concept Graph

As a microblogging service, Twitter is playing a more and more important role in our life. Users follow various accounts, such as friends or celebrities, to get the most recent information. However, as one follows more and more people, he/she may be overwhelmed by the huge amount of status updates. Twitter messages are only displayed by time recency, which means if one cannot read all messages, he/she may miss some important or interesting tweets. In this paper, we propose to re-rank tweets in user’s timeline, by constructing a user profile based on user’s previous tweets and measuring the relevance between a tweet and user interest. The user interest profile is represented as concepts from Wikipedia, which is quite a large and inter-linked online knowledge base. We make use of Explicit Semantic Analysis algorithm to extract related concepts from tweets, and then expand user’s profile by random walk on Wikipedia concept graph, utilizing the inter-links between Wikipedia articles. Our experiments show that our model is effective and efficient to recommend tweets to users.