基于作者引用文献关系的潜在研究兴趣主题发现

iscovering the authors' research interests is vital for detection of future research topics and fast sharing of science papers in a certain field. When researchers write academic papers, they will cite some literatures, which can reflect the authors' research interests to some extent. Thus, the citation relationship can be used to discovery authors' potential research interest topics. By extending LDA (Latent Dirichlet Allocation) model, this paper proposed an author-interest topic-paper model focused on citation relationship. In this model, each author is associated with multinomial distribution over all the topics, while each topic is associated with multinomial distribution over all the cited literatures, based on which the latent potential research interest topics can be detected. Experiment on citation relationships from DBLP dataset showed that it can discover potential interest topics and representative cited literatures effectively. Besides, an author's interest distribution on different topics can also be described.