Topic Enhanced Sentiment Spreading Model in Social Networks Considering User Interest

Abstract

Emotion is a complex emotional state, which can affect our physiology and psychology and lead to behavior changes. The spreading process of emotions in the text-based social networks is referred to as sentiment spreading. In this paper, we study an interesting problem of sentiment spreading in social networks. In particular, by employing a text-based social network (Twitter) , we try to unveil the correlation between users' sentimental statuses and topic distributions embedded in the tweets, then to automatically learn the influence strength between linked users. Furthermore, we introduce user interest to refine the influence strength. We develop a unified probabilistic framework to formalize the problem into a topic-enhanced sentiment spreading model. The model can predict users' sentimental statuses based on their historical emotional status, topic distributions in tweets and social structures. Experiments on the Twitter dataset show that the proposed model significantly outperforms several alternative methods in predicting users' sentimental status. We also discover an intriguing phenomenon that positive and negative sentiment is more relevant to user interest than neutral ones. Our method offers a new opportunity to understand the underlying mechanism of sentimental spreading in online social networks.