Effects of event sentiment on product recommendations in a microblog platform

The occurrence of a planned event often closely correlates with human emotions. Since words describing an emotion are important vehicles for capturing an event, to understand how emotions affect subsequent product recommendations, in this study, we use the microblogging platform “Plurk” in conjunction with a Chinese emotional lexicon developed previously. More specifically, we identify emotional keywords as a basis for distinguishing user sentiment in event-based posts; in the meanwhile, we use these emotional keywords as a gauge for product recommendation. Further, we utilize T-test, one-way ANOVA, and two-way ANOVA approaches to explore the effects different emotional variables have on product recommendations. Our experimental results show that under the influence of positive events, negative emotional posts of a microblog user can enhance the product recommendation effect better than the positive posts themselves. In product recommendations with various emotional words, positive emotional recommendations are able to achieve better recommendations than that of either negative or neutral recommendations. The results of our study can effectively help service and product suppliers carry out social marketing during a given event period and utilize emotional variables to effectively attract user interest in clicking on advertisements on social network sites, thereby achieving the goal of enhancing the product recommendation effect.