Sentiment-based User Profiles in Microblogging Platforms.

ABSTRACTTwitter has become one of the major platforms for self-expression in the Social Web, mostly due to its adoption by mobile users and its short message format. This presents endless possibilities for social behavior researchers that, for the first time, have access to massive amounts of data generated by humans. Nevertheless, most of the current research on emotions in social platforms focuses on reactions to particular events, or crowd behavior. In this article we present our research in the identification and characterization of user sentiment profiles in online social media. By analyzing a dataset of more than 36,000 users, we identify several distinctive groups, according to similarities in their sentiment behavior. We study differences and similarities between these profile clusters and present detailed statistics. We found that a large number of Twitter users can be grouped in nine distinct profiles according to the strength and polarity of their sentiment. Researchers and practitioners can benefit from our approach to characterize Twitter users in several scenarios, such as social recommendation, and mood estimation.