Research Of Product Opinion Targets Extraction From Websites Of E-Commerce

A large number of product reviews are generated in the Internet with the rapid development of e-commerce. The information is rich in content and reliable which provides purchasers valuable reference and help. User opinion information is so much and natural that artificial reading cannot get an accurate and comprehensive understanding of the products. So opinion mining research for reviews from the Internet emerged.Opinion mining includes topic extraction which is also called opinion targets extraction, holder identification, claim selection, sentiment analysis, etc. Opinion targets extraction focuses on the identification of the product attributes and characteristics from the reviews of the e-commerce websites. It is the basic of opinion analysis and opinion summary which brings great convenience to users for reading reviews quickly.Methods based on rules and templates are not effective in invalid comment sentences. Besides, some opinion targets of low frequency are usually ignored. As the problems above, a three-level filter method for opinion targets extraction is proposed which contains bootstrapping candidates extracting, association confidence computing and domain confidence computing. In the first level, a set of opinion target seeds is adopted to extract candidate opinion targets and opinion words from opinion texts. In the second level, the association between the opinion target and opinion word is used to estimate the association confidence of every candidate opinion target and candidate opinion word. High association confidence makes high accuracy. In the third level, an uncorrelated domain is adopted to calculate the domain confidence of every opinion target in the rest set which is for mining the opinion targets of low frequency. The experimental results on three public datasets demonstrate the effectiveness of the proposed approach.