Design And Implementation Of Product Analysis System Based On Opinion Mining

With the rapid development of Web2.0, a large number of customer reviews for products or service emerge in the e-commerce website, professional forums and many other websites. These user reviews contain a wealth of information, which not only lead to a dramatic change on the business processes, but also bring a profound impact on the consumer’s behavior. There is growing evidence that the reviews influence consumer’s purchase decisions. Through the analysis of product reviews, we can dig out the main features of these products, and find out users’opinions and attitudes towards them. This allows sellers to better understand customers’feedback and adjust their business strategies, and makes customers capable of finding the specific features information without reading a great number of reviews, and helps them make right decisions.However, the online reviews have the characteristics of rapid formation, powerful interaction, as well as randomness and variability of language. Analyzing and processing these huge amounts of reviews with uneven quality by manual method will be a very tedious and time-consuming task. Moreover, it is difficult to obtain all the useful information from these reviews. Therefore, an unstructured data mining technology-"Opinion Mining", which is based on the goal of obtaining the useful information from all the product reviews, attracts more and more attention from scholars.Therefore, in order to adapt to the Internet environment, the paper selected dianping.com as the study object. We eventually collected more than20000online reviews of food stores in Yunnan Province by using crawl tech, and we also developed a system to process these reviews. In this paper, we mainly analyze the reviews of food shops and then make some research on the extraction and clustering of product features and sentiment analysis. The specific study of this paper are as follows:.1) Present an approach to extract product features, which is based on Apriori algorithm, and using PMI with the seed set and co-occurrence degree with opinion words to filter features.2) Present an approach to group product features based on K-means algorithm, in which sharing words, lexical similarity and opinion words are chosen as the tokens to represent the association of product features.3) Present an approach to analyze the sentiment polarity of product reviews based on semantic analysis and vector space model.In order to verify the effectiveness of the proposed method in this paper, we conducted some experiments on extraction and clustering of product features and sentiment analysis. Reviews on food store from dianping.com were analyzed as experimental data. And as can be seen from the experimental results, the method proposed in this paper is effective and also has good engineering application value. At the end, the paper is summarized and the further research idea and direction is prospected.