Opinion Mining Based On Product Reviews And Study Of Sentiment Analysis

With the rapid development of the Internet, all aspects of social life are undergoing the tremendous change, especially the rise of the mobile Internet which is making it more convenient and frequent to participate in networking activities, particularly prominent is the behavior of the user reviews during online shopping. Because the product reviews contains a wealth of valuable information, the effective mining of the reviews of the product information are great significance to guide business to improve product quality and to provide a reference for consumers. In this paper, the study focused on the recognition of evaluation object, the recognition of the evaluation word, the extraction of the relationship between evaluation object and evaluation word and emotional orientation analysis in Opinion Mining.For the recognition of the evaluation object and the evaluation word, based on the theory of the dependency syntax, the paper proposes the iterative identification algorithm of evaluation elements based on the dependency syntax. Because of a direct dependence relationship between the evaluation object and evaluation word, the evaluation object and evaluation word are in a pair. So by extracting those pairs in the corpus, then recognize the evaluation object and evaluation word through iterative identification.In the extraction of evaluating collocation, this paper proposes the passing algorithm based on the relationship of dependency syntax. Based on the theory of dependency syntax, the dependency syntactic tree formed in the sentence elements and the relationship of dependency syntax among the tree nodes, in the way of traversing the dependency syntactic tree, we can extract evaluation objects、evaluation words and modifiers in which there are exist dependencies, and then get evaluating collocation and filter the error one in the candidate set of evaluating collocation in the support of the area knowledge base, thereby improving extraction effect. In the emotional analysis of evaluating collocation, with the emotional expression patterns, in accordance with the evaluation of the expression pattern of the emotional vocabulary and the syntax mode, this paper analysis the emotion of the evaluating collocation through the dictionary-based approach.Finally, based on the above research, the paper designs and implements a reputation analysis system based on online product reviews. Accessing to the specific areas of product reviews through the web crawler technology and in the support of the corresponding field of knowledge, with the use of the opinion mining methods of this study, mining the valuable information contained in the product reviews.