Analyzing and Detection of Human Emotion through Online Social Networks

Abstract

Online social networks play an important role for sharing the emotion between users. It has become increasingly prevalent in various applications, including personalized ad services and recommendation systems. However, the traditional approach to emotion analysis has focused exclusively on sentence-level polarity, ignoring the information that multiple emotions can coexist in users' minds. In this study, it provides deals with the problem of multiple emotions from the user's view, which formulates a multi-label learning problem. Through analysis of an annotated Twitter dataset, we identify correlations between emotion labels, social correlations, and temporal correlations. Based on the findings from the state of art techniques a factor graph-based emotion recognition model that incorporates social, temporal, and social correlations with emotion labels. Our model utilizes a multi-label learning strategy and can detect multiple emotions more accurately than existing baselines. Overall, our study provides a novel approach to emotion detection in OSNs, with potential applications in personalized services and recommendation systems