GSLDA: Supervised topic model with graph regularization

In this work, we study the problem of regularizing supervised topic model using graph structure. Supervised topic model generates each document independently, whereas in many applications there are links among documents, which are quite useful for refining topics. To overcome this limit of supervised topic model, we propose a regularization framework using graph structure. By leveraging both textual content and link structure, the output of the proposed model can promote effect of topic extraction and social network analysis simultaneously. Experiment results on two real datasets demonstrate the effectiveness of the proposed approach.