Research On Sentimental Polarity Multi-Classification For Online Social Networks

In recent years,with the rapid development of Internet technology and the increasing popularity of intelligent mobile terminals,online social networks have developed rapidly.As representative of online social networking platforms,Weibo and Wechat have become indispensable parts of people's daily life.By online social networks,people are free to express their own sentimental attitudes towards various foods,movies,events,and so on.Due to these behaviors,a large amount of textual data with sentimental attitudes is generated.By collecting and analyzing these textual data with sentimental attitudes,it can provide data support for applications such as public opinion analysis,hotspot analysis and objection evaluation.And it will make relevant analysis more accurate and further facilitate people's lives.The analysis of textual data with sentimental attitudes mainly relies on the sentiment classification technology in natural language processing.In order to make the result of sentiment classification more refined,this paper studies the sentimental polarity multi-classification problem in sentiment classification.In order to better evaluate the effectiveness of the solution for sentimental polarity multi-classification problem,it is necessary to develop comprehensive evaluation indicators.Based on the various evaluation indicators applied to the classification problem in machine learning,this paper combines them with the specific sentiment multiclassification problem,and put forwards the evaluation indicators applied to this problem.These evaluation indicators can comprehensively reflect the performance of the model that solves sentimental polarity multi-classification problem in all aspects,and they would alleviate the problem of imperfect evaluation indicators of sentimental polarity multi-classification.Meanwhile,these evaluation indicators are also used to evaluate and analyze the effect of our solution in the subsequent research.Based on the evaluation indicators,this paper studies the sentimental polarity multiclassification in combination with the data set for movie reviews.Data is the support of research,and it will have a large impact on the performance and effectiveness of the model.This paper analyzes the data set that is commonly used in the problem of sentimental polarity multi-classification,and proposes the data enhancement method to improve the data set.By this way,a larger and more balanced data set is established.In order to verify the validity of our method,experiments are performed on the improved data set.The experimental results show that after applying our classification model and the improved data set for training,the overall classification accuracy rate of 52.3% is reached on the sentimental polarity multi-classification problem with five sentiment categories.The result is 1.7% higher than the effect on the original data set,and the classification effect between different categories is more balanced.At the same time,the other evaluation indicators proposed in this paper are almost improved,and the effectiveness of our method is verified.