Mining Product Feature And Opinion Based On Pattern Matching

Review mining means extracting the users’ evaluation information from network reviewsto help consumers carry out rational consumer and provide the necessary referenceinformation for enterprises to enhance competitiveness. Feature words and opinion wordsmining is a very important research topic. It is the basis of the review mining. This paper hasa research on feature words and opinion words mining.A mining method based on pattern matching is proposed in this paper. First constructmatching patterns by using part of speech combination, to create the conditions for extractingfeature words and opinion words. Then identify the users’ points from the comments throughpattern matching, and extract feature words and opinion words in pairs. Through patternmatching we can not only identify the users’ points which Part of Speech (POS) of opinionwords is adjective, but also identify the users’ points which POS is verb. So the problem thatthe traditional method didn’t consider the opinion words which POS is verb is solved.Extracting both feature words and opinion words effectively reduces the situation that thefeature words and opinion words didn’t match each other in the result set, and enhance boththe recall and precision of the result set. In addition, pattern matching in mining users’ pointsdid not consider the occurrences of the feature words, so both the frequent feature andinfrequent feature can be mined.In order to make the users’ points mined more accurate, this paper also does research onturning complex sentences which traditional method did not think over. On the basis of theanalysis the structure characteristics of turning complex sentence, an improved SBValgorithm is proposed, which extracts the user point by making use of the SBV relationshipand structural feature of the transitional complex sentence.The experiments show that both the pattern matching method proposed in this paper andthe improved SBV algorithm are effective.