Emotion Modeling from Writer/Reader Perspectives Using a Microblog Dataset

Most recent studies on emotion analysis and detection focus on how writers express their emotions through textual information. In this paper, we model emotion generation on the Plurk microblogging platform from both writer and reader perspectives. Support Vector Machine (SVM)-based classifiers are used for emotion prediction. To better model emotion generation on such a social network, three types of non-linguistic features are used: social relation, user behavior, and relevance degree, along with textual features. We found that each of the non-linguistic features can be combined with linguistic features to achieve higher performance. In fact, the combination of linguistic, social, and behavioral features performs the best.