The Study Of Viewpoint Recognition And Sentiment Analysis Based On Topic-typed Microblog Comments

With the popularity of the Internet in China, people’s life styles have changed gradually, and anincreasing number of people get and release information via the Internet. Netizens begin to spreadtopics on politics, sports and entertainments. A large number of microblog comments which worthanalyzing are generated every day. These comments contain a lot of emotional information andideas.Here come a large number of studies on microblog. Through analyzing the studies of Chinesemicroblog sentiment, it is found that Chinese microblog sentiment classification methods can bedivided into two categories, one is the method based on emotion lexicon, another is the methodbased on machine learning. Due to the complexity of Chinese grammar, the method based onmachine learning can’t retain the emotional relationships between features, and doesn’t have theability to deal with the situation of many emotional words and evaluation objects. When using themethod based on emotion lexicon to classify the sentiments, the less consideration of thevocabularies of microblog topics and the distinction between different emotional words leads to lowaccuracy of classification results. The previous studies on Chinese sentiment analysis are lack oftargeted research and the influence of evaluation objects to emotional polarity and strength. Theprevious extraction algorithm of emotional features needs to be improved.To solve these problems, this paper uses the method based on emotion lexicon, and selectChinese emotion word ontology as a basic emotion lexicon, which can effectively improve thedisadvantage of previous emotion lexicon not distinguishing emotional strength. Furthermore, thispaper adopts a computation method of word’s semantic similarity bases on HowNet, constructs anfield emotion lexicon. Considering the influence of evaluation object on microblog sentimentclassification, this paper constructs a reasonable evaluation object lexicon. This paper adopts rulesand SVM model to recognize viewpoint, selects comments related to microblog topic, and improvethe quality of comment text in sentiment analysis. Then we conduct some reprocessing on themicroblog text, combine smoothing technique and grammar rules as the emotional featureextraction method of microblog comment text to process the negative words, adverbs of degree,microblog emotional signs, emotional words and evaluation objects in microblog comment text.Finally this paper uses emotion calculation formula to classify the emotional tendency of comments,effectively improve disadvantage of previous Chinese microblog sentiment analysis, achieve anreasonable calculation formula to judge the emotional strength of microblog comments. Differentmicroblog topics have different field lexicon. This paper constructs targeted emotion lexicon and evaluation object lexicon based on topics to have a specific topic emotion analysis, which canimprove the effect of sentiment analysis.This paper uses microblog corpora provided by Data Tang as experimental data, including thedata in the fields of life, traffic accident, and science and technology. The experimental results showthat the proposed model of viewpoint recognition and sentiment classification has betterclassification effect when compared with previous classification model. The experiment verifies thevalidity and rationality of the proposed method.