User Sentiment Analysis Of Chinese Social Platform Based On Text Mining

As the number of social platforms in China continues to grow,Internet users are increasingly inclined to express their opinions on social platforms.These opinions tend to have emotional orientation,positive tendencies,neutral tendencies,and negative tendencies.Positive and negative emotional tendencies of the specific topics to find the topic of netizen center and explore what users have positive and negative feelings on the content,as well as the evaluation of these content,the platform has a very important guiding significance.At present,the three well-known social platforms in China are "Douban","Zhihu" and "Jianshu" platforms respectively.If text data are crawled separately from each platform,not only the workload is huge,but it is easy to be generalized,Therefore,this article chooses to crawl text data from the third-party platform "Weibo" search keywords,the data of this article is obtained by searching Douban,Zhihu and Shuji respectively in the search bar of Weibo.In order to study the rigor,In this paper,LDA theme model is used to summarize the original text data to verify whether the original data crawled can represent the situation of each platform.At present,emotion classification of texts is divided into two categories:classification based on sentiment dictionary and classification based on statistical model.In this paper,classification based on sentiment dictionary is adopted.Firstly,based on the existing authoritative open-source sentiment dictionaries,the basic sentiment lexicons and negative lexicons are constructed and filtered,and then the online sentiment lexicons are added manually.The text uses the original data of artificial emotional polarity marked as the test corpus,and separately uses the constructed emotion dictionary and the statistical model-based classification method(SVM)respectively to judge the polarity of the test corpus.The results show that the affective dictionary based classification method effect it is good.Next,the constructed sentiment lexicon is used to make an emotional analysis of the crawled text data to analyze the emotional polarity of the users' opinions on the platforms.Followed by positive and negative emotional orientation of the text keyword analysis to do to find out what users were dissatisfied with or support.These contents have great guiding significance to each platform.