NLP-Based Research On Opinion Mining Of Chinese Web Review Sentences

Opinion mining and polarity analyzing is the process of automatically identifying opinion and classifying polarity with the application of computer technologies, in which topics or features that users show interest in will be extracted and semantic orientation and subjective strength will be computed.This paper focuses on the sentimental analysis of Chinese review sentences. It proposes a novel algorithm for computing the contextual polarity of polar words. And a framework is introduced for topic identification and feature extraction, providing an innovative solution for the association of extracted opinion to a specific topic which is the most challenging task for opinion mining. We finally compare the performance with the results of manual annotation. The experiment has shown the algorithms are both reasonable and effective.For the semantic orientation of words, past researches consider only their prior polarity, i.e. static polarity. It would not only result in misjudge of the word's polarity, but also affect the precision of opinion mining. This paper introduces methods to compute word's static polarity, and also analyzes its dynamic polarity by tracking its contextual environment.For opinion mining, this paper introduces syntactic parser to approach the most challenging problem of mapping modifier to its corresponding subject.