The Impact Of Sentiment Extremity And Inconsistency Between Rating And Sentiment On Helpfulness Of Online Hotel Consumer Reviews

The exponential increase in the number of online reviews makes it difficult for consumers to deal with information overload.Determining which factors make reviews more helpful for consumers to make decisions in the purchase process has become a focus of research in the field of online reviews.Although there are rich research achievements in the field of helpfulness of online reviews,the research direction of text sentiment is still to be further studied.This paper examines the impact of textual sentiment extremity and the inconsistency between rating and sentiment on the helpfulness of online reviews in the hospitality industry and explores the moderating effect of text information quality characteristics(text topic and review length).In this paper,a multivariate regression model was established based on the hypotheses,and more than 700,000 online reviews from 655 hotels in New York City of TripAdvisor were selected as data source.Naive Bayes classification model a nd LDA subject classification model were used to calculate the text sentiment score and topic probabilities respectively,as required variables.On this basis,this paper conducts empirical analysis on the effects of text sentiment extremity and the inconsistency between rating and sentiment on helpfulness of review respectively.The results of the hypothesis verification are as follows:(1)text sentiment extremity has significant positive effect on helpfulness,and extreme negativity is more helpful than extreme positivity.That is,the relationship between text sentiment and review helpfulness is asymmetric U-shape.(2)the number of text topics has a positive moderatio n effect,while type of search attribute topics has a nagative moderation on the effect of sentiment extremity on helpfulness,compared with type of experience attribute topics.(3)the inconsistency between rating and sentiment has a significant negative effect on helpfulness,and the effect of high score inconsistent reviews on helpfulness is stronger than that of low score inconsistent reviews.The helpfulness ranking of the four types of reviews is: low rating & negative sentiment > high rating & positive sentiment > low rating & positive sentiment > high rating & negative sentiment.(4)review length negatively moderates the impact of inconsistencies on helpfulness.Thus,compared with inconsistencies in short texts,consumers are less likely to be negatively affected by inconsistencies in long texts.The research of this paper has certain significance to theory and practice.The research from the perspective of text sentiment extremity and the inconsistency between rating and sentiment,fills the gap in the research field of the helpfulness of online reviews,and the results can help e-commerce platforms better manage reviews.