Sentiment Analysis Based On E-Commerce Commodity Review Text

Massive e-commerce commodity evaluation information contains great commercial value.At the same time,a large number of e-commerce commodity review information has caused some troubles for manual screening and selection.Therefore,how to use efficient sentiment analysis technology to enable computers to automatically classify e-commerce product evaluation information for shoppers or commodities.Production companies provide more accurate comments and emotional information,help users to quickly improve product reputation,correctly select products,and help companies improve product qualityThe main purpose of this paper is to conduct sentiment analysis of Internet commodity reviews.The main research work includes1?Pre-processing work for product reviews.Select the mobile phone comment on the e-commerce website as the research object,use the crawler to collect the data,and perform the text pre-processing on the obtained comment data,including the text screening filter,data cleaning,Chinese word segmentation,part-of-speech tagging,and delete the stop words.Etc.Preparing for sentiment analysis of subsequent review texts2?Feature selection and feature weighting.First of all,feature selection plays a decisive role in sentiment classification Selecting appropriate features is beneficial to improve the effect of sentiment classification.Through the analysis of various feature selection algorithms,this paper makes appropriate improvements based on the Information Gain(IG)algorithm,and adds the intra-and extra-class frequencies of features as an indicator of feature selection to the calculation of information gain values.Secondly,for feature weights,this paper is based on the TF-IDF algorithm,and also adds the intra-and extra-class frequencies of features as an indicator of the feature weights calculation.Experiments show that the improved feature selection and feature weighting algorithm improve the accuracy of classification.Finally,the two improved algorithms are combined to complete the feature selection and feature weighting,and used in the subsequent sentiment classification.3?Multi-decision combination classification model construction.Based on three decision-making ideas,a multi-decision combination classifier is proposed.The support vector machine classifier is used to make the first three-branch decision for the text set.For the generated boundary domain I,the kNN algorithm is used for the second three-branch decision.For the boundary domain II,the naive Bayes algorithm is used for the third time.The decision is made for the boundary domain III by the Naive Bayes classifier,kNN classifier and the support vector machine classifier weighted voting.Experiments show that the mu lti-decision combination classifier helps to improve the correct ra te of sentiment classification and has certain superiority.