Research On Topic Sentiment Analysis Based On Weibo Public Opinion Event

In recent years,with the development of economic level,people’s quality of life has generally improved,mainly in various aspects such as food,clothing,housing and transportation,etc.Meanwhile,with the rapid development of the current Internet,a more open and shared information era has arrived,and people can express their opinions in real time on the Internet as a way to express their emotions.Microblogs are loved by the public for their fast dissemination and no threshold,and have become the most timely and influential online opinion dissemination platform.It is of great practical significance to monitor public opinion by collecting the comments of netizens on microblogs and grasping the changes in topics and emotions of netizens’ attention to public opinion events.This paper mainly uses the method of combining topic model and emotion analysis to divide the topic and analyze the emotion of the comment data under microblog related topics.Using Python custom programming,it crawls Internet users’ comments starting with the microblog topic "Hongxing Erke donated 50 million to help Henan Province".According to the changes in Internet users’ discussion and event popularity,it divides the public opinion stage and analyzes the emotional trend and theme changes of Internet users in each stage.In terms of research methods,this paper firstly mines the topics under different stages with the help of LDA model to clarify the topic hotspots that netizens pay attention to under different stages of public opinion,so as to lay the foundation for the subsequent sentiment analysis.When using sentiment analysis,the original sentiment dictionary cannot be fully applied because of the rapid update of Internet terms and the use of emoticons in comments also represents a certain intensity of emotion,so we need to build a sentiment dictionary exclusively for this microblog opinion event.Firstly,we select emotion seed words and emoji seed words,and further apply the SO-PMI algorithm to find other words containing emotion tendency by combining a certain number of seed words,and compare and extend the obtained emotion words with the original basic emotion lexicon,so as to further improve the accuracy of emotion recognition.The experimental results show that the method proposed in this paper can effectively take into account the co-occurrence and semantic association among words,which can improve the performance of sentiment analysis and make the sentiment score more accurate.Meanwhile,the exclusive domain sentiment lexicon is used to finegrained mining of microblog comments,analyzing the proportion of sentiment categories under different opinion stages and grasping the sentiment trend of public opinion events as a whole.To sum up,this paper combines the stage theory of public opinion,divides public opinion events into different stages,analyzes the comments under different stages by LDA model for topics,constructs a sentiment dictionary exclusive to this public opinion event,makes more accurate judgments on positive and negative sentiment tendencies and sentiment categories in microblog comments,reveals the changes of topics and sentiment trends that netizens pay attention to,and grasps the characteristics of the evolution of public opinion under the overall topic events.