Research Of Sentiment Analysis For Internet Commodities Review

In the booming e-commerce network environment, more and more subjectivity comment text about products appears on various shopping websites. These comments in the text include users’ emotional tendencies towards various aspects, such as love, hate and so on. Sentiment analysis on these comments can help businesses keep abreast of the advantages and disadvantages of products, improve product quality a nd services. It also can provide data support for the potentia l consumers to make decisions. Sentiment analysis technology(Sentiment Analyzing) takes advantage of these vast amounts of text comments, digs out the user’s preference of emotional information technology. More and more researchers are involved in the research work in this field.The main task of sentiment analysis is to mark positive or negative emotions a user express toward a product from the given text. The content of the study includes the subjective and objective content identification, the classification of the emotional tendency, the calculation of the emotional strength and so on. It involves multiple research in natural language processing, machine learning, text classification etc. The main research focus of this paper is to classify the positive or negative emotions expressed by the subjective text.This paper proposes a clustering algorithm based on property of co mbination neural network with comments of products, and classify the property of products with this algorithm. Then the text is presented as a four-dimensional vector representation method and combined with the SVM algorithm to achieve the sentiment analysis of product reviews. Cyber language often appears in comments of inter net products, they can also express positive or negative sentiment. In view of this characteristic, this paper proposes method to build dictionary of comment emotion of products based on Google’s word2 vec, and analyze comment text with this method.Combination neural networks based automatic clustering method considers the positional relationship between attribute words and context of the words. It clusters the related attribute in comment text according to grammar and context. By clustering, comment text is divided into several smaller clusters, each cluster marked with a category label art ific ially. Comments text in each category label is for the property of products. Since product comments has its own characteristics, this paper converts comments text into a four-dimensional vector, compares it with doc ument frequency and information gain feature selection algorithms using real data set of comments of products, classifies emotional tendency of comments of products with SVM algorithm, verif ies the accuracy and validity of the method. By training the word2 vec tool, we build emot ional dictionary of comments of products, then classify emotional tendency of comments text with this dictionary, experiments proved that this method has higher classification accuracy.