Fuzzy Inference Research Based On Consumers' Sentiment Of Online Reviews

Sentiment analysis of online reviews is one field that has received more and more interests, and its goals are to analyze reviewer's attitude or sentiment, furthermore to mine potential useful business information. With the development of Electronic Commerce online reviews become the richest interactive resources between consumer and business,consumer and consumer. Natural language texts of online reviews implicit a number of information of consumers'complicated perception,emotion,willingness. However, few researches have focused on sentiment analysis and inference from the perspective of natural language fuzziness. So this paper constructs fuzzy semantic resources of online reviews, fuzzy emotive lexicon ontology and fuzzy emotion corpus. Based on them, using consumer behavior and psychology theory constructs fuzzy inference rules and do inference researches.About constructing semantic resources, firstly, the paper present the definition of the fuzzification method, using Basic Emotion and Evaluation Fuzzy Set to process the polarity and intensity of lexicon. Through the combination of manual classification and acquiring the intensity automatically the work of corpus tagging is fulfilled of lexicon particle size fraction. Then the definition of Fuzzy Operator of Modifiers is given and the Privative Dictionary and Intensifier Dictionary are constructed. Lastly, calculation method of phrase and sentence emotion value is provided.On the procedure of inference research based on consumer psychology and behavior theory, first of all, the calculation method of inference antecedent is put forward, combined with the preference weight of product attributes and fuzzy calculation method the product comprehensive evaluation and emotion value is offered. For the calculation of emotion classified multi-dimensionally according to the status of the emotional classification and the features of consumers'emotion of online reviews the classification system is determined. By the Basic Emotion Fuzzy Set and fuzzy matrix positive and negative emotion value are obtained. During the procedure of inference rules bases construction, on the one hand for consumers with different consumption motives recommendation inference of product with different emotion and evaluation intensity is realized; On the other hand, according to the relationship between perception value fairness and consumer positive or negative emotion consumer perception value fairness is inferred. Experiments showed inference method based on consumers'emotion of online reviews is reasonable. These two methods explore inference by consumer psychology and behavior from diverse point of view.