Which Tweets Will Be Headlines? A Hierarchical Bayesian Model for Bridging Social Media and Traditional Media

Microblogging platforms such as Twitter provide a convenient channel for people to express their feelings, report news, and communicate with friends. Most existing work on social media analysis has been focused on predicting users' behaviors, analyzing the corresponding social networks, tracking the popular topics, etc. However, there is limited research effort on uncovering the relationships between social media (e.g. Twitter) and traditional media (e.g., Washington Post and New York Times), which has a big impact in our daily lives and our society. This paper targets on a novel and important research problem as which and whose tweets are favored by the traditional media. The basic intuition is that whether a tweet could be picked up or not by traditional media depends not only on whether its content matches traditional media's interests towards this specific user but also the writer's personal influence, reflected by factors such as the number of followers. Based on this intuition, this paper proposes a Twitter Pick-Up Relational (TPUR) model to simultaneously integrate these factors. In particular, the dependence between the traditional media's interests towards a user and the content of each tweet, and the influence of each user are integrated in a hierarchical bayesian model. An extensive set of experiments are conducted on two datasets from two popular microblogging platforms, i.e., Twitter and Sina Weibo (Chinese version Twitter), to demonstrate the advantages of our algorithm against baseline methods on the proposed problem.