Fine-Grained Sentiment Analysis Of E-Commerce Online Reviews

With increasing growth of reviews and commodity attributes becoming more diverse, the overall emotional tendency already can’t satisfy the needs of buyers and manufacturers. Consumers hope to get analysis results of target attributes and manufacturers hope to master strengths and weaknesses of products and also consumers’tastes. In this situation, the traditional sentiment analysis failed to help consumers reduce cognition cost and information overload in a more fine-grained level-attribute level, and manufacturers can’t further understand the advantages and disadvantages of a product and develop a niche market. Moreover, some reviews do not directly point out features in the statement which we need to understand according to the context of comments.Due to the traditional sentiment analysis could not refine consumer tastes and clear the strengths and weaknesses of online products, this study carries on a fine-grained sentiment analysis research which takes as an example that a women’s dress in taobao.com. Based on the analysis of the previous research about attribute recognition and classification, this paper puts forward a method which finds explicit and implicit attributes separately and for attribute classification we make full use of morphemes in words. Implicit features identification is a very tough task, however, we take advantage of POS rules to extract a large number of field implicit attributes. And this paper establishes a common field emotional dictionary in order to make up for the inadequacy of previous researches which always overly depend on existing emotional dictionary. At the same time, this paper designs a comparison algorithm when calculating sentiment strength and proves the analysis of implicit attributes gets the accuracy of sentiment analysis raising.Through the above process, we get the visual result of fine-grained sentiment analysis and illustrate how it could help customers and manufacturers make decision. We have the following contributions in this paper:In theory, implicit features identification is a very tough task, however, we take advantage of POS rules to extract them; In application, discovering a niche market from implicit features and this is significant to manufacturers whose marketing based on long-tailed market online. At last, by the general methods we use to crawl reviews, induce general category of product attributes in women’ dress and find implicit features based on POS rules, we increase generality of the application.