Research On E-commerce Review Helpfulness Based On Topic Modeling

In recent years,with the popularity of the mobile Internet and the rapid development of e-commerce,people have become accustomed to the consumption of daily life through the Internet.When consumers conduct shopping activities on the e-commerce platform,they often refer to other purchasers.Commentary information is used to make purchasing decisions,and previous studies have shown that consumers are more willing to trust online reviews than the product information displayed by merchants.User reviews on e-commerce platforms have become an important reference for people making purchasing decisions about products or services.However,an important issue that arises is that as the number of reviews on the e-commerce platform increases and the quality of reviews is mixed,it is difficult for online merchants to understand the true concerns of customers about their products or services.It also makes reviews those are helpful for consumer decision making obscured with huge amounts of useless commentary.This paper takes two of the 10 stores in the popular public review website as examples.We propose Help-LDA,Tfidf-LDA,Max-LDA and classic LDA models to analyze the topics and helpfulness of a total of 4,000 online reviews.The experimental results demonstrate that the Help-LDA model can help merchants to interpret reviews,and Help-LDA+SVM is helpful in commenting helpfulness prediction.On the one hand,the Help-LDA model can help merchants to understand the real concerns of online consumers about goods or services,and to improve the quality of goods and services of the merchants.On the other hand,it can explore the relationship between the topic of reviews and the helpfulness of reviews,to predict new reviews for purchase decisions.The topic extraction from e-commerce reviews in the paper can provide great managerial implications for both online merchants to understand and online consumers in online shopping decision making.