The Predictive Power Of Discrete Emotions In Online Reviews Over Sales

With the rapid development of the Internet and the popularity of online review platforms,the online review is playing a more and more important role in E-commerce business,profoundly influencing consumer's decision process and seller's marketing strategy.The sentiment is an important aspect of the online review,attracting great attention of platforms,consumers,and sellers.Recently,domestic restaurant industry is growing quickly.However,high rivalry is among the existing players in the industry as well.Sales prediction hence is of great importance to the sellers in this industry.Existing studies have been using valence(the positive/negative polarity of sentiment)to predict or explain sales.However,valence is not enough to fully depict the subtle and intricate nature of human sentiments.Discrete Emotion Theory,on the contrary,could better delineate human sentiments by using distinct discrete emotions.Thus,this work aims to explore the predictive power of discrete emotions in the online restaurant reviews.This study used Word2 Vec model to automatically construct an emotion lexicon specific to online restaurant review domain and proposed a methodology of using emotional vectors to represent discrete emotions in online restaurant reviews from dianping.com,the largest domestic online review platform.Then,this study used valence,discrete emotions,historical sales and review volume data to construct sales prediction models.We compared the predictive power of valence and discrete emotion,and found that the predictive power of discrete emotions is statistically significantly higher than valence.The contributions of this study are as follows.First,we originally proposed a complete methodology for automatic domain-specific lexicon construction,and for analyzing discrete emotions from online reviews.Second,the lexicon constructed by this study is also a contribution.It could help researchers and sellers to better delineate emotions embedded in the online restaurant reviews on dianping.com.Third,this study enriches the existing sales prediction literature by using discrete emotions as predictive variables,and we proved that they can enhance the prediction accuracy.Finally,this study empirically found discrete emotions could better depict human sentiments than valence.