Impact of Text Diversity on Review Helpfulness: A Topic Modeling Approach
In this study, we aim to investigate the impact of an important characteristic of textual reviews – the diversity of the review content on review helpfulness.
Consumer-generated reviews are an essential format of online Word-of-Month that help customers reduce uncertainty and information asymmetry. However, not all reviews are equally helpful as reflected by the varying number of helpfulness votes received by reviews. From consumers’ perspective, what kind of content is more effective and useful for making purchase decisions is unclear.
We use a data set consisting of consumer reviews for laptop products on Amazon from 2014 to 2018. A topic modeling technique is implemented to unveil the hidden topics embedded in the reviews. Based on the extracted topics, we compute the text diversity score of each review. The diversity score measures how diverse the content in a review is compared to other reviews.
In the literature, studies have examined various factors that can influence review helpfulness. However, studies that emphasized the information value of textual reviews are limited. Our study contributes to the extant literature of online word-of-mouth by establishing the connection between the diversity of the review content and consumer perceived helpfulness.
Empirical results show that text diversity plays an important role in consumers’ evaluation of whether the review is helpful. Reviews that contain more diverse content tend to be more helpful to consumers. Moreover, we find a negative interaction effect between text diversity and the text depth. This result suggests that text depth and text diversity have a substitution effect. When a review contains more in-depth content, the impact of text diversity is weakened.
For consumers to quickly find the informative reviews, platforms should incorporate measures such as text diversity in the ranking algorithms to rank consumer reviews.
Future study can extend the current research by examine the impact of text diversity for experienced goods and compare the results with search goods.