Research On Fine-grained Sentiment Polarity Classification Of Chinese Network Consumption Review

In recent years, with the simultaneous rapid development of E-commerce and various types of internet interactive community, network consumerism review has formed big data sets which have significant commercial value. The sentiment analysis technology of network consumerism review can identify and extract related tendentious information from subjective text, to quantify the description of subjective opinion which is expressed in text through specific algorithms, thereby to infer the preferences degree of reviewers to the goods or services purchased. It can influence the decision-making behavior of the consumers and also has an important guiding significance to improve products push accurate product information.In order to make the utility of sentiment analysis a more precise role in a property, to maximize the commercial value of consumer review, this article conducts researches on the automatic extraction algorithm of emotional words and attribute words in network consumer review analysis, and fine-grained network consumer review emotional polarity classification based on the analysis of sentiment words and attribute words. Firstly, this article crawls the network catering consumerism review as corpus using web crawler technology, by virtue of the theoretical knowledge in areas such as machine learning, natural language processing, combining the characteristics of Chinese network comments language expression, proposes an automatic construction method of field sentiment lexicon based on corpus; Then this article combines the characteristic analysis of sentiment words and attribute words in catering consumerism reviews, proposes two kinds of attribute words search strategy respectively based on noun+LDA and word spacing+LDA+emotional thesaurus, then verifies the effectiveness of methods and search strategy by experiments and results analysis; this article constructs sentiment evaluation unit, and conducts sentiment computing according to the sentiment evaluation unit, realizes fine-grained text sentiment analysis on the basis of the calculation score, the fine-grained sentiment analysis results are input displayed in form of graphically hierarchy.The research of this article can further improves the accuracy of user reviews sentiment analysis, play a greater guiding role in the product or service improvements and the consumer decision-making process.