Analysis Of Emotional Tendency Of Online Comments Based On Improved SO-PMI Algorithm

Emotion is a special attribute of human beings.The content of this paper is the sentiment analysis in the user comment text in the e-commerce environment.Comments can reflect the user's feelings about using the product,and the user can express appreciation and vent their dissatisfaction.In addition,comments are a major reference for other users to purchase the same product.However,because of the large number of comments,the quality is uneven and the content is different,it is difficult to accurately classify,which is inconvenient for the user's browsing.Therefore,the sentiment analysis method is introduced from the emotional orientation,and the comments are classified and optimized to help the user judge the quality of the product through comments,so that the user can determine the needs efficiently and accurately.This article takes the user's purchase of mobile phone as a research example and uses sentiment analysis to judge the emotional tendency of the comment.The first step is to construct an accurate and comprehensive sentiment dictionary for the mobile phone field,use the improved SO-PMI algorithm to expand the emotional words,and further,based on the sentiment dictionary,calculate the emotional words and other categories of words,and judge the emotional tendency of the comments to classify it.The main research contents are summarized as follows:(1)Basic emotion dictionary merging and auxiliary dictionary tidying.In sentiment analysis based on sentiment lexicon,due to the limitation of vocabulary and version in traditional single sentiment lexicon,the sentiment analysis of specific domain comments is less effective.Therefore,by collecting and sorting,I choose Emotional Dictionary of Hownet,NTUSD and the Li Jun Chinese Bao-Bian Emotional Dictionary of Tsinghua University to combine,so that the general basic sentiment dictionary was obtained.The dictionary of negative words,the dictionary of degree adverbs and the dictionary of related words were also collected.Finally,considering that the specific object of this article is the mobile phone,a mobile web dictionary is added to comment on the participle and build the basic sentiment dictionary.(2)Emotional dictionary expansion.The mobile phone commentary also contains words that are not added in the basic sentiment dictionary,affecting the effect of sentiment analysis.To judge the emotional tendency of mobile phone reviews containing new words,it is necessary to expand the emotional dictionary.To this end,this paper uses the improved SO-PMI algorithm to expand the sentiment dictionary.The main improvements include:1.The selection of benchmark words was optimized using the TF-IDF algorithm.2.Based on the polynomial NB classifier,the optimal classification dimension is determined,and then the chi-square statistic is used to extract the candidate emotional words.(3)Mobile online commentary emotion classification experiment.After considering the data volume,word of mouth and product quality of the shopping platform,choosing to crawl the comment data of the mobile phone section of Jingdong Online Mall.After de-duplicating and formatting the comments,the comments with annotations are obtained,and then the classification is tested based on the sentiment dictionary.The results show that the sentiment classification method can effectively classify the online comments of mobile phones.The effectiveness of the dictionary expansion method based on the improved SO-PMI algorithm is proved by experiments,which provides a new idea for the expansion method of the emotion dictionary.The precision,recall and F1 value of the verified classification reach a high level.Finally,according to the classification results of the improved SO-PMI algorithm,a comparative experiment of affective tendencies was designed to classify the positive comments and negative comments.The positive comments and negative comments with the highest affective tendencies had the highest classification accuracy.