The Product Review Mining Key Problem Research Based On Feature

With the rapid development of Internet, especially the rise of Web2.0, more and more people choose to go shopping online, which to the extent drives the development of E-commerce. Meanwhile the further development of E-commerce also brings people greater convenience for shopping online. Previously before shopping, people acquainted themselves with products in advance only depending on public praise and advertisement effect. But now, manufacturers provide consumers with review function and they can state reviews on products after experiencing products for a while. At the same time, the subsequent consumers also look through these review information before they go to buy the products online. With a very large number of product reviews online, consumers can’t look through review information piece by piece. If we dispose of these plentiful review data by hand, then this is a very busy and time-consuming task. So, the product review mining technology is born at the right time.The so-called product review is a kind of document that people express their opinions on product’s parts, property or performance. As a research branch of text mining, product review mining has gained rapid development and a parts of results has been applied to practice. Product review mining, involving several fields including natural language processing, information retrieval, information extraction, data mining and machine learning, has been one of research hotspots from academia. The main content of the thesis is as follows:We give a common frame and list the research tasks of product review mining. The research tasks can be divided into three main tasks including subject and object analysis, sentiment classification and product review mining based on fine granularity. According to three tasks, we introduce the present status of field accordingly. In addition, we also introduce the status of field on product review mining prototype system.The subject evaluation relation identification is one of tasks from product review mining. The article focuses on subject evaluation relation identification of Chinese product review mining. The differences between Chinese and English and the research on English product review mining prior to Chinese product review mining, makes it impossible to transplant the technology and method on English product review mining into Chinese product review mining directly. So, the research on Chinese product review mining run into tremendous challenge. Therefore, we improve product feature-opinion word pairs based on review mining of Chinese dependency grammar. Specially, first, we set several extraction rules using Chinese dependency grammar. Secondly, we make use of these rules to extract candidate product feature-opinion word pairs. At last, we prune mismatched product feature-opinion word pairs by feature rank and named entity identification system.In the end, we conduct experiments and experiments data comes from microphone and digital camera. We give experiment results of product feature and product feature-opinion word pairs under different threshold and analyze them. The experiment results show that our method achieves desired effect.