Research On Sentiment Classification Of Chinese Micro-blog Text Based On Three-way Decisions

With the rapid development of mobile Internet,the network information exchange platform represented by the micro-blog is widely used.The Chinese micro-blog text contains a lot of valuable information,you can understand the people's attitudes and opinions about an event or a product by classifying it emotionally.The result of the micro-blog text sentiment classification is very important and valuable in consumer research,market analysis and public opinion monitoring.How to quickly classify positive and negative emotional information from the micro-blog text is the starting point of the study.In recent years,the use of classification technology and NLP technology for emotional classification of the micro-blog text has become a hot topic of research.This paper uses the sentiment classification of Chinese micro-blog text as the main line,and proposes a semi-supervised microblog text sentiment classification method based on three-way decisions,and the research work is as follows:(1)This article studies the selection of emotional features and weight calculation methods.High-dimensional features in sentiment classification can cause redundancy,the commonly used feature selection methods only consider the role of emotion feature words in the category without considering their emotional features,coupled with the sparse and irregular features of the micro-blog text,resulting in the effect of the micro-blog text sentiment classification often not ideal.In order to improve the accuracy of the micro-blog text sentiment classification,this paper proposes a method based on Z-score emotional feature selection.Experiments are conducted to compare the classification effect of the micro-blog text sentiment classification using CHI,VCHI and Z-score feature selection algorithms,and the effectiveness of the micro-blog text emotion feature selection method based on Z-score is verified.(2)The classification of subjective text for Chinese microblog has been researched.In the emotional classification,the unsupervised learning method often has low classification accuracy,the supervised learning method must be able to obtain the ideal classification effect under the condition of marking a large amount of corpus,which takes a high labor cost.Semi-supervised learning method can make full use of unlabeled corpus information to improve the classification accuracy in the case of a small amount of corpus.This paper introduced three-way decisions theory into the micro-blog text sentiment classification,made full use of the advantages of three-way decisions in the classification of uncertain objects,and combined the advantages of R-self-training algorithm in dealing with the corporal imbalance problem,a semi-supervised micro-blog text sentiment classification method based on three-way decisions is proposed.Experiments show that the proposed method in this article can effectively improve the accuracy of the micro-blog text sentiment classification in the case of less corpus.