Emotional Comment Text Analysis And Its Application In The Credit Assessment

Internet has evolved from a static information carrier into a free speech platform. E-commerce is a new type of economic activity in the network, the semantic orientation and opinion of comments on its web site has become an important basis for decision making and the most convincing propaganda. People are generally more willing to believe the recommendations and suggestions provided by other consumers, so almost all e-commerce sites provide the consumer reviewing platform to help consumers understand the products, and help businesses master consumer feedback. However, with the rapid development of the Internet economy, review resources rapid expand and the cost of accessing and understanding of the required information from the massive review text grows exponentially. How to take advantage of these massive emotional text resources and analyzing subjective content of text has become a research hotspot.Domestic and foreign research data shows that sentiment analysis technology has a wide range of applications, which can not only satisfy people’s demand of identification and analysis of subjective information and reduce the searching and understanding cost of users, but also contribute to the discovery of new knowledge. However, the problems like the diversity and complexity of the natural language expression, the lack of standardization of network language and the field dependence of sentiment words gave sentiment analysis set the challenge to sentiment analysis. Moreover, it is difficult to confirm the information’s credibility based on the similarity of the sources due to the network information coming from anonymous individuals. The credit problems like false description of the product, fake and shoddy have constrained the development of e-commerce. Therefore, the study of text sentiment analysis and e-commerce credit evaluation model based on review sentiment analysis has important practical significance and economic value of developing a trusted e-commerce.The main works of the paper are as follows:1. The paper researches the theories and methods of sentiment analysis, proposed a text vector model based on emotional feature according to the characteristics of Chinese reviews. We combine the semantic information with traditional text vector model successfully by using the six-tupel (object, opinion, over-modifier, general-modifier, negation, punctuation) as text vector feature. The model adequately considers a variety of factors which affect the emotions of the text and effectively link sentiment word to reviewed object and its context to confirm the semantic orientation and strength of the text.2. We propose a novel weighting algorithm for sentiment analysis of Chinese based on the proposed vector model. Analysis of the use of adverbs of degree can reflect the impact of different vector feature to semantic orientation of the text. Experimental results demonstrate that our proposed HADC method can effectively improve the accuracy of review text sentiment classification.3. A sentiment feature extraction method based on dependency parsing is proposed in the paper. Developed a corresponding recognition rules and achieve better results in identification of sentiment word, evaluation object and modifiers.4. We analysis the problem of e-commerce credit evaluation system and build the credit evaluation indicator system and model. Research and design of sentiment analysis method, compute sentiment distribution of users’ comment and extreme degree of sentiment polar, define credit evaluation indicators based on the results of sentiment analysis, and combined AHP which combines qualitative analysis with quantitative analysis to define credit evaluation indicator system.We give full consideration to the mutual influence among the credit of the consumer, the credit of the comment and the credit of the businesses to construct an improved credit evaluation model which contains the transaction amounts and transaction time. The experiment shows that the improved model is more comprehensive, objective and scientific in assessing credit of both trading parties.