Microblog, with its wide participation and convenience, has changed the way that people get news about current events. In recent years, lots of breaking news and hot topics are released by microblog platform firstly, as well as its much wider and more distribution than traditional media platform. Extracting these useful information in real-time will help us to grasp the latest and hottest topics which currently discussed by microblog users. However, due to the micrblog content is short and sparse, traditional topic extraction methods can not be used directly. In this paper, we propose a new topic model named Microblog Features Latent Dirichlet Allocation (MF-LDA) to extract microblog topics. We incorporate five microblog's unique features: support, comment, retweet, publish time and user authority into LDA model. These features are utilized to compute each microblog's attention value, authority value and word frequency. And the higher feature value of a term, the greater probability of a hot topic it be. Experimental results on real datasets demonstrated our MF-LDA model is more efficient and accurate than other methods in hot topic extraction.
展开▼
