Emotional Analysis Based On Product Review

With the rapid development of the Internet and e-commerce,more and more consumers express their views and opinions on the website.Such opinion information contains an emotional expression of something,service,or business.Collecting and analyzing such emotional information is of great value to business organizations and even to national governments.In general,we put the shopping left after the attitude of the business or product evaluation called product reviews.As a source of emotional information,it is widely used by individual consumers and business organizations.This dissertation takes the product review as the main research object,studies the question of the reliability and emotional polarity of the emotion data in the emotion analysis.In order to reduce the noise of the emotion analysis datasets and filter the spam comments,the first part of this dissertation proposes a spam comment recognition method based on random forest and rule matching.Use of a variety of features of the effective integration,combined with random forest algorithm,rule matching and other technologies to identify spam.In order to verify the effectiveness of the proposed identification method,a large number of qualitative and quantitative comparative experiments were carried out under different commodity-based datasets.The experimental results show that the accuracy of this method is above 80%.For the credibility of emotional datasets,the second part of this dissertation proposes a false comment recognition method based on distributed outliers.This dissertation analyzes and summarizes the important features of false comments on the Internet,builds a feature model of product reviews based on multidimensional feature space,and more effectively labels false reviews by adding statistical indicators such as the anomaly of word distribution and abnormal value of emotional distribution.Finally,the classification optimization algorithm Co-Bagging is used to improve the recognition accuracy of false comments.Experimental results show that the method has higher accuracy,recall rate and F value under the same premise.Further,in order to more accurately judge the emotional orientation of the product reviews,the third part of this dissertation proposes a method based on emotional dictionary to analyze the emotional orientation of the product reviews.Methods Aiming at the characteristics of emerging online emotion words in the product reviews,the Chinese emotional polarity dictionary HowNet is used as the basic dictionary and the word2 vec tool is used to construct the sentiment evaluation dictionary of commodity comments automatically through semantic similarity calculation.In addition,aiming at the calculation method of emotional sentiment in Comment text,the SO-PMI-DR algorithm is improved.The experimental results show that the sentiment classification of the product reviews can be effectively improved through the automatically constructed sentiment dictionary and the improved SO-PMI-DR algorithm.