Detecting Hotspot Information Using Multi-Attribute Based Topic Model.

摘要: Microblogging as a kind of social network has become more and important in our daily lives. Enormous amounts information are produced shared on basis. Detecting hot topics the mountains can help people get to essential quickly. However, due short sparse features, large number meaningless tweets other characteristics microblogs, traditional topic detection methods often ineffective detecting topics. In this paper, we propose new model named multi-attribute latent dirichlet allocation (MA-LDA), which time hashtag attributes microblogs incorporated into LDA model. By introducing attribute, MA-LDA decide whether word should appear or not. Meanwhile, compared with model, applying attribute gives core words an artificially high ranking results meaning expressiveness outcomes be improved. Empirical evaluations real data sets demonstrate that method is able detect accurately efficiently several baselines. Our provides strong evidence importance temporal factor extracting