Retweet Prediction in Sina Weibo Based on Entity-Level Sentiment Analysis

Abstract

Sina Weibo and other microblogs have rapidly become significant means by which people communicate with the world and each other. There has been a large number of studies surrounding retweet behavior, focusing on a variety of features that may explain why people retweet certain weibo content or certain user’s weibo. However, they didn’t consider the sentiment aspect or simply use the sentencelevel sentiment analysis which is not accurate enough. In this paper, we use Wikipedia and syntactic infomation to build user profile with entity-level sentiment analysis, then use this profile to predict user retweet behavior and with other common features we try to figure out what role entitylevel emotion plays in user’s retweet bahavior. Experiments on real weibo dataset shows our approach’s effectivceness and emotion is an important factor in retweet behavior.