A HIERARCHICAL FRAMEWORK FOR ONLINE PRODUCT REVIEW HELPFULNESS ASSESSMENT

Abstract

Online product reviews have become valuable resources to facilitate consumers in assessing the products, to support manufacturers in comprehending consumer opinions, and to help retailers in enhancing consumer loyalty. However, the high volume and high variation in quality has brought challenges for knowledge acquisition. Although many websites allow consumers to vote for the helpfulness, the approach is limited in its efficacy because many websites do not provide this voting function and most reviews receive very few votes. Various factors (e.g. review length and readability) have been explored for review helpfulness assessment; however, most of them have focused on superficial characteristics of review text while ignoring its deep-level semantics. This dissertations aims to address above limitations by proposing a hierarchical framework for product review assessment. This dissertations makes multifold contributions: first, a rule-based method for extracting product features from online product reviews has been proposed, which requires no training data; second, a hybrid bottom-up method is developed for the construction of product feature hierarchy, and a hierarchical framework is created to help understand review helpfulness; third, drawing on the hierarchical framework and product uncertainty and information quality theories, we propose a research model that consists of three novel main factors ? breadth, depth, and redundancy. Last, we test our model with both experience goods and search goods, which demonstrates the generality of our approach. In addition, our framework and model offer new lens for explaining mixed findings about impacts of review length and rating on review helpfulness. We conduct experiments with online reviews of different types of products to test the proposed techniques and research model. The findings provide strong evidence for the importance of the hierarchical framework for improving our understanding of review helpfulness. This research provides several implications. First, this dissertations applies and extends the product uncertainty and information quality theories to the domain of online product reviews, which guides the design of factors for explaining review helpfulness. Second, the dissertations operationalizes the factors by constructing a hierarchy, which provides a showcase on how to utilize semantic relations among product features to gain deep insights into online reviews.