Research On Micro-Blog Sentiment Classification Based On Data Expansion And Emotion Analysis

The Micro-Blog platform can provide massive data,and Micro-Blog platform data updates faster and attracts a large number of researchers.The sentiment analysis of the text on the Micro-Blog platform is also changing rapidly.Quality and authoritative Chinese text data sets which marked are few,it leads underperform on the sentiment analysis.In general,researchers only analyze events and opinions in text to predict user's sentiment,but ignore the users' emotion feature,so the practical application of sentiment analysis technology has been effected.Due to the annotated authoritative Chinese data sets are relatively few,the text information that can be extracted and analyzed is less,so the prediction accuracy is relatively low.The paper puts forward DESA(the Data Expansion Sentiment Analysis Algorithm).The DESA algorithm expands corpus through sentiment dictionary,thesaurus dictionary,and antonyms dictionary firstly.For positive and negative data,the paper uses the dictionary of antonyms to replace sentiment words;for neutral data,the paper use the thesaurus dictionary to replace sentiment words.In the process of sentiment analysis,in addition to considering the sentiment polarity of the original data text,the paper expands the sentiment polarity of the new text to jointly predict the sentiment polarity of the original data.Experimental results show that data expansion effectively improves the accuracy of sentiment analysis technology.Since the traditional sentiment analysis technology only analyses the events and opinions,but ignores other information of the text.The paper puts forward ETC(the Emotion Text Classification Algorithm).The ETC algorithm gets Chinese dictionary of emotion to check stress word and relaxation word in the text,combines with the classification results of DESA,and predicts sentiment polarity from the original corpus.The experiment shows that ETC has better experimental effect.