Multi-Source Emotion Tagging for Online News Comments Using Bi-Directional Hierarchical Semantic Representation Model

With the rapid growth of news services, users can now actively respond to online news by expressing subjective emotions. Such emotions can help understand the preferences and perspectives of individual users, and thus may facilitate online publishers to provide users with more relevant services. Research on emotion tagging has obtained promising achievement, but there are still some problems: Firstly, traditional methods regard a document as a flow or bag of words, which cannot extract the logical relationship features among sentences appropriately. Therefore, these methods cannot express the semantic of the document properly when there exists logical relationship among the sentences in the document. Secondly, these methods use only the semantic of the document itself, while ignoring the accompanying information sources, which can significantly influence the interpretation of the sentiment contained in documents. In order to solve these problems, this paper proposes a hierarchical semantic representation model of news comments using multiple information sources, called bi-directional hierarchical semantic neural network (Bi-HSNN), which not only captures the sentiment among words in sentences, but also applies a bottom-up way to learn the logical relationship among sentences in the document. This paper tackles the task of emotion tagging on comments of online news by exploiting multiple information sources including comments, news articles, and the user-generated emotion votes. A series of experiments on real-world datasets have demonstrated the effectiveness of the proposed model.