Research Of Products Review Mining And Summarization

With the recent proliferation of Web2.0sites and applications, people have witnessed the revolution that Internet brought to us. E-commerce is one of the products of the Internet and people’s shopping habits is deeply changed by its popularity. Nowadays, many of the e-commerce websites are not only a place to do trading business, but also a place for people to make their reviews of some products. Online customer reviews are considered as a significant informative resource which is useful for both potential customers and product manufacturers. They can be referenced by other customers when they have trouble in deciding which product to buy. They can also be used by manufactures to get feedback about their products for product development, market research and consumer relationship management.However, with the steadily increasing volume of e-commerce transactions, the amount of user provided product reviews are increasing rapidly on the web, some popular products can even get thousands of reviews at some large merchant sites. It becomes more difficult for users to read all of the relevant review documents. Demand has thus been growing for opinion mining techniques that can automatically analyze user reviews from large quantities of written data and extract the most desired information for users.Based on this background, the paper studies the problem of generating feature-based opinion summarization of customer reviews, mainly to complete the work in the following areas: (1) We proposed an improved method based on Apriori algorithm to extract main features of a product. The algorithm of PMI is used to filter the features, making a progress with the precision of feature extraction. At the same time, a method base on machine translation model is used to extract the implicit features.(2) For the identification of sentiment orientations of the review, a set of fine-grained and stratified scoring formulae is designed using part-of-speech tagging, grammatical dependencies and word sentiment scores.(3) As to the method of generating a summarization, a structured form of feature based summarization is proposed in order to show the overall information of the review efficiently.In order to verify the effectiveness of the method proposed in the paper, some test is designed and the result seems good. At the end, the paper is summarized and some future research idea and direction is prospected.