Research On User Generated Content Emotion Analysis Method

User generated content generally refers to the emotional analysis by the dictionary based on semantic rules or machine learning method based on active web users to generate content mining and analysis,and get some subjective information expression of attitudes and opinions from users,such as user for an event point of view,suggestion,mood,emotion.Nearly two years,with the "Internet +" action continues to help the development of e-commerce business,more and more attention has been paid to the study of the emotional analysis of electronic commerce evaluation information.This study is based on a Taobao commercial text evaluation information as the research object,through the analysis of a commodity evaluation information,so that consumers and businesses understand the real information and user experience of goods from different angles,then provide the decision-making basis for consumers to buy the goods,also allows businesses to continuously improve product quality,improve products and services and the appropriate sales plan.This paper use the similarity calculation method of PMP and "knowledge network" to assign the corresponding weight to the emotion words,based on the semantic rules,multi classification machine learning and fuzzy comprehensive sentiment analysis method for each evaluation object from a commodity value,favorable rate,analysis of the product evaluation information for multi angle comprehensive commodity value,And verify the effectiveness of the three methods.The main work is as follows:1Emotional analysis of commentary information based on the weighted dictionary.First of all,on the basis of the basic emotional dictionary,combined with the "knowledge network" similarity calculation and statistical PMI method to give a certain weight to the word,with the weight of the domain of emotional dictionary,in addition,the construction of negative words,degree adverbs,Turn the word emotional dictionary and give different weights.Then,the evaluation information tuples are obtained,and the emotional extremum of the commentary sentences is obtained by the corresponding rules of the emotional polarity calculation,and the emotional sentences are classified according to the emotional extremes.Finally,the validity of the emotional dictionary with the weight field and the rule of the affective computation for the emotion analysis are verified.2A Machine learning approach based on multi-classifier for sentiment analysis of review information.The text representation,feature selection and weight calculation method were introduced,and the characteristics of evaluation for short text information improved the method of calculating Boolean function feature weights,weights of words instead of "1","0" of text representation,and use of the parameter to reduce the weight of similar evaluation information distance,to improve the accuracy of classification.And feature selection was used ^2 statistical method,through KNN,SVM and naive Bayesian methods three classification models verify the effectiveness of the improved method of feature weight calculation of Boolean functions.3Emotional analysis of review information based on fuzzy comprehensive evaluation method.According to the need of fuzzy comprehensive evaluation,the index system is constructed,and the index of evaluation object should be belonged to,and the method of emotion analysis is introduced.Then the fuzzy statistical method is used to determine the weight of the fuzzy comprehensive evaluation matrix and the index.Finally,the comprehensive evaluation value of the goods is obtained.Fuzzy comprehensive evaluation method is used to evaluate the evaluation information of a product.