Contextual Polarity and Influence Mining in Online Social Networks

In this paper, we combine social networks and sentiment analyses to identify significant and informative community divisions. We further demonstrate analytically that positive sentiment between individuals increases by the number of discussions exchanged. We employ our method on real-world social traces of Starbucks ideas and propose the models of correspondence among the idea proposers and the authors of the comments. Our results reveal that sentiment scores can indeed lead to more informative and signicant community structure improvement over the social networks without sentiment scores. The results also show that there is a statistically significant quadratic relationship between the sentiments scores and the number of conversations between users.