Multi-Features Group Emotion Analysis Based on CNN for Weibo Events

With the development of social network, group emotion analysis on social media such as Facebook, Twitter and Weibo becomes a new trend in recent years. Many different methods have been proposed for group emotion analysis, including traditional methods like SVM and NB and deep learning methods like RNN and CNN. This paper proposes a CNN model with multi-features. We first analyze the characteristic of weibos to collect features including basic features, user-based features and content-based features. We introduce these features to our CNN model to analyze emotions for Weibo events, which has been proved in experiment that it is effective to get the accurate sentiment of weibos. We crawl 5,319,687 weibos about 724 events from Sina Weibo and apply several feature matrices to classify them to 4 types. We use the new model to analyze 4 kinds of events to get the group emotion. The results suggest that the classification we suggested can capture the emotions within different event group.