Microblog Text Sentiment Analysis Research

As a kind of social application with timeliness,activity and rapidity that other media can't match,Microblog has a large number of users.Microblog has become an important carrier of information dissemination in various fields such as major events,citizen rights and social assistance.However,Microblog allows users to express their opinions anonymously and has insufficient supervision.Therefore,there are many false information,negative speeches and content that undermines national unity in Microblog.If this bad information spreads over a large area,it may form a network public opinion,which is easy to cause political,economic and social contradictions.Therefore,emotional analysis of Microblog can understand the views of netizens on some hot events in real life and predict future development trends,so that the government can control in time and guide public opinion to develop in a positive direction.The microblog text contains a lot of online buzzwords,pictures,emoticons and other content,which makes the effect is not good when performing sentiment analysis.Moreover,personality affects the user's writing and expression.People with the same personality traits tend to choose similar emotional expressions,but they rarely consider personality in the process of emotional analysis.Therefore,this paper proposes a character-based microblog sentiment analysis model PLSTM.Main tasks as follows:1.Established a classification rule for personality.According to the analysis of the Big Five model,the characteristics of different personalities are obtained and the corresponding personality dictionary and expression dictionary are established.Each personality dictionary contains common words under the personality.The corresponding personality judgment rules are formulated and the text satisfying a certain personality is divided into the character set,and one text can conform to multiple personality judgment rules,so one text can also belong to multiple character sets at the same time.2.Using Word2 vec to vectorize the text,use the LSTM(long and short memory neural network)to train the emotional classifiers of different personalities,use the classifier and emotionally classify the text,and output the probability value of the text emotional category.Then,using the bagging integration method to integrate the output of multiple basic sentiment classifiers,the integrated combination method uses the weighted sum,and finally the emotional tendency of the text.Through experimental comparison and analysis,the results show that the combined personality and integrated learning sentiment analysis model PLSTM can improve the accuracy of microblog text sentiment analysis.Explain that it is effective to conduct targeted emotion classification for different character sets and to conduct integrated learning.