| Human basic emotions are independent,unique and can be distinguished from each other,and in terms of Discrete Emotion theory of psychology,all human emotions are viewed as a combination of these basic emotions.Based on such theory,some early researches in online English reviews obtained notable results.By contrast,it is just initiated to apply discrete emotions in online Chinese reviews,especially in the field of consumption.With regard to distinctive Chinese people’s emotional expression,this research investigates a large quantity of Chinese reviews on mobile phone products at Jingdong Mall(www.jd.com),a large-scale e-commerce platform in China by adopting different theories,models and technical means of management,computer science,psychology,human behavioral dynamics,linguistics and some other disciplines.The research includes three main aspects:recognizing different discrete emotions in Chinese online review,concluding distribution pattern and exploring its impact on online sales.Natural language processing technology and supervised machine learning method are used to classify the discrete emotions implied in online reviews in accordance with OCC model of psychological affective cognitive evaluation theory.After comment-tagged corpuses with six discrete emotions being constructed,including satisfaction,disappointment,admiration,reproach,love and hate,sub-tagged comment sets with six discrete emotions are made feature selection based on Chi-square statistics(CHI).And five classifiers based on Bayesian(GaussianNB),K-nearest neighbor classification(KNN),Logistic regression(LR),Random Forest(RF)and Support Vector Machine(SVM)are used to classify discrete emotions.The results show that the classifier based on Logistic Regression(LR)algorithm has the best classification effect on each discrete emotion,and the classifier based on different algorithms has different classification effect on the same discrete emotion.This study poses a new idea and methods for emotional recognition of online Chinese comments,and makes a reference to other researchers engaged in Chinese online commentary sentiment analysis.(2)Relying on statistical methods and human behavior dynamical researches,the study finds the distribution patterns of consumers’ discrete emotional expression based on discrete emotional classification of massive online Chinese commodity reviews.The findings show how review stars,price ranges and "purchase-comment" time intervals distribute on collective level in terms of the attributes of commentaries,commodities and time.The reviews varying from one to five stars suggest the percentage of reviews expressing love and hate alters the most.The research also includes instances whatever prices mobile phones are,which wouldn’t impact the consumers who are always willing to express their love and satisfaction in the reviews.The more moderately products are priced,the more value consumers attach to their performance price ratio,but for commodities with the high price,consumers are more likely to evaluate the services provided by merchants,the third-party e-commerce platforms and logistics,etc.The frequency curve of"purchase-comment" time interval shows "bimodal distribution".The first peak of different discrete emotional reviews all occurs on the first day after purchase,with the highest frequency of release of emotional reviews containing admiration and reproach,and the second peak occurs on the eleventh day after purchase,especially love,hate and satisfaction.The six discrete emotion reviews follow the power law distribution at different"purchase-comment" time intervals,and the power exponents corresponding to three positive discrete emotions are larger than those corresponding to three negative discrete emotions.The time intervals of "purchase-comment" of different emotional comments are obviously paroxysmal,among which the paroxysmal index values of admiration and reproach are the highest while the paroxysmal index values of hate emotion is the lowest.The memory index for the "purchase-comment" interval of six different discrete emotional reviews are close to zero,indicating that each consumer’s purchase behavior and comment behavior are respectively individual behaviors on the e-commerce platform.To a certain extent,this study identifies the deficiencies of the existing research on Sentiment Analysis of online reviews,which is rarely explored in the law of emotion distribution.It enlarges researching scope of emotional information of online reviews and points a new research direction for the behavior characteristics and academic findings of user generated content.(3)The study uses Appraisal-Tendency Framework Theory(ATF),Persuasion Effect and the Two-process Theory to examine how the six discrete emotions implied in online reviews effect the online sales of commodities.From the perspective of behavior orientation,the regression model,a method of management science,is constructed to analyzes the internal mechanism between different discrete emotions and potential consumers’ purchase behaviors.The study finds that discrete emotions with the same valence implied in online reviews do not have the same impact on online sales of goods.Among the three positive discrete emotions,only admiration plays a significant positive role in online sales;among the three negative discrete emotions,only hate has a significant negative impact on online sales.The brand competitiveness has a regulatory effect on-the impact of discrete emotions on sales.The proportion of reviews containing hate has a greater impact on the sales of goods of weak brands than strong brands.The study is different from the current situation in which most of the researches on discrete emotions implied in online reviews apply the questionnaires or experiments simply aiming at exploring the perceived usefulness of different discrete emotions.And the study helps to well understand specific emotions in online reviews and expands the related research on online word-of-mouth. |
