Abstract: Product reviews presented on the Web by the customers are used to be as the feedback on products and meanwhile providing shopping experience for potential customers. However, with the increasing of commodity review information, consumers tend to be submerged in these comments. This paper presents an algorithm for product reviews mining base on opinion mining. First, we collect product reviews using crawler. Second, identify the product features which the customers mention in the view. Third, find the opinion sentences. Last, compute the emotional scores of the opinion sentences and output the emotional distribution. The aims of researching on this question are to provide consumers with more clear, more detailed evaluation to goods.

Key words:  commodity comments, opinion mining, affective computing, watershed algorithm