Reasearch On The Emotional Analysis Of The Evaluation Texts For The E-commerce Website

With the continuous development of Web2.0 technology,more and more users began to comment and share experience on the Internet,The text contains the emotion of the user,and the text sentiment analysis technology has become the focus of current research.In the face of the Internet environment change rapidly,with the rapid growth of information,we need to find a suitable method to obtain the required content from a large amount of text information,so as to cope with the challenge of information explosion.It is a difficult point in the field of Natural Language Processing to classify the comment text in the information and get the emotional information.The traditional text sentiment analysis,mainly for the text level and the sentence level text,such as stock analysis,public opinion monitoring etc..However,with the deepening of the research,the user has put forward higher requirements to the emotion analysis technology.The traditional coarse grained sentiment analysis technology can not meet the needs of the times.In this paper,the traditional methods of emotion analysis are improved,and the practicability and accuracy of emotion analysis are improved by exploring new ideas.In this paper,we propose a new method to construct Co-word matrix for sentiment analysis.Through the "co-occurrence" of word pairs,we solve the problem of "high frequency word isolation".By calculating the semantic similarity and sentence similarity of word pairs,we can solve the problem of "homogeneous quantity" in Co-word analysis.By combining these two methods,the Co-word matrix is constructed,which improves the accuracy of clustering.In this paper,a new method of constructing emotional dictionary is put forward according to the characteristics of Internet information.On the basis of HowNet evaluation dictionary and emotional dictionary,the emotional words in the comment information are extracted,and the negative word list,the word list and the degree adverbs are constructed.For the quantification of emotional words,this paper does not use the traditional method,the quantitative evaluation of emotional words,but through the review of the existing score information,based on semantic analysis technology to build different models.From the point of view of the fuzziness of emotional words,we calculate the membership degree of emotional words for each segment,and finally get the emotional value of the emotional words,so that the quantification of emotional words is more reasonable and effective.