Analysis On Sentiment Evolution Of Network Public Opinion For Weibo

Social networking platforms such as Weibo have become an important information communication platform due to their low thresholds,information flow,and graphic expressions.On Weibo,users do not only share their lives,but also express their sentiment about popular social incidents.Analyzing sentiment evolution of social incidents by massive Weibo users is crucial for opinion monitoring,branding,and knowledge management.This thesis studies the problem from two perspectives.(1)Public sentiment evolution in social incidents.A new public sentiment evolution model is proposed which assumes that background sentiment evolve in the public’s sentiment evolution of social incidents,and the background sentiment keep changing smoothly and slowly.In order to improve the effect of the public sentiment evolution model,this paper also proposes an entity-level sentiment polarity detection algorithm to improve the accuracy of sentiment polarity detection on Weibo comments.Experiments on real data sets show that our proposed method is better than the contrast method used in modeling the emotional evolution of social events.(2)Sentiment prediction.A multi-task learning model is proposed which assumes that the ability of news to trigger public sentiment shift is mainly related to the sentiment feature and evidence feature contained in the news,and if the news triggers public sentiment shift,it is mainly reflected in the news will cause the number of comments to change and the sentiment polarity of comments to change.By applying our model to large sets of news and tweets,we demonstrate its significant improvement over baseline methods.