Multimodal Deep Belief Network Based Link Prediction and User Comment Generation

摘要

In social network services, the relationship among members can be represented as link network, and the link prediction problem is to infer the value of such links. By analysing the structure of link network, researchers have proposed several methods for solving link prediction. Nowdays, when some members label their link values, they also make comments, which have been seldom considered for link prediction. In this paper, by considering both the link network data and user comment data, we propose multimodal deep belief network based link prediction method, which outperforms other state-of-art methods. With the learned joint distribution of link network features and user comment features, our method could generate comment words properly.