Analysis And Research On Sentiment Tendency Of Short Texts Of Online Reviews

In recent years,with the rapid development of social media and e-commerce,there are a large amount of valuable comment short texts information has been continuously generated on these social network sites and e-commerce sites.By effectively analyzing and judging the sentiment tendency of these comment short texts,we can not only obtain important key information from them,but also further promote the development of related fields such as e-commerce,public opinion analysis.Firstly,we built an expanding sentiment lexicon based on the sentiment corpus.Secondly,we proposed a short text sentiment tendency analysis method based on integrating sentiment contribution.Finally,we implemented a system for analyzing the sentiment tendency of short comment text.The main work of this article included three aspects which were sentiment lexicon building,sentiment tendency analysis and system implementation:(1)Proposed a method of building sentiment lexicon based on integrating the sentiment relevance of categoriesIn this article,we firstly built a basic sentiment lexicon by integrating the universal Chinese sentiment lexicon,and based on this,we selected reference sentiment words from the sentiment corpus.Secondly,considering that text categories affected the results of word sentiment tendency discrimination,we combined negative words,frequency of words,and Mutual Information method to calculate the sentiment relevance between candidate sentiment words and text categories.Thirdly,considering that reference sentiment words affected the results of word sentiment tendency discrimination,we combined reference sentiment words to calculate the sentiment relevance between candidate sentiment words and reference sentiment words.Then,on this basis,we implemented a word sentiment tendency discrimination method based on integrating sentiment relevance of categories,and integrated the sentiment words identified by the proposed method with the basic sentiment lexicon to form an expanding sentiment lexicon.Finally,we verified through experiments that the method had a certain improvement in the macro average F1 value and the extended expanding sentiment lexicon had better classification performance.This further showed that the proposed method was feasible in word sentiment tendency discrimination and the constructed expanding sentiment lexicon was practical in text sentiment classification.(2)Proposed a short text sentiment tendency analysis method based on integrating sentiment contributionFirst of all,considering that sentiment words affected the result of the sentiment tendency of short text,we combined the category conditional probability of sentiment words and the semantic contribution weight of sentiment words to calculate the sentiment contribution of sentiment words.Secondly,considering the clauses that affected the result of the sentiment tendency of short text,we calculated the sentiment contribution of clauses by combining the position information of the clauses and the number of sentiment words it contained.Thirdly,considering that the number of different polarity sentiment words affected the result of sentiment tendency of short text,we used the sentiment words with the same sentiment tendency as a sentiment whole,and determined the sentiment contribution of sentiment whole based on the number of sentiment words.Then,based on this,a short text sentiment tendency method based on integrating sentiment contribution was realized.Finally,we verified through experiment that the method had a certain increase in macro average F1 value,which further proved that this method was valid in judging the sentiment tendency of short text.(3)Implemented a system for analyzing the sentiment tendency of short comment text based on integrating sentiment contributionThe system could not only collect short text data of the latest reviews of related products on Jingdong Mall website,but also identify subjective comment short texts from these data.Then the proposed short text sentiment tendency analysis method was used to further analyze the sentiment tendency of these subjective comment short texts.Finally,the intuitive sentiment tendency analysis results were displayed to the user,so as to further provided relevant reference for the users in purchasing goods.