Identifying helpful online reviews: A product designer's perspective

  • Authors:

  • Ying Liu;Jian Jin;Ping Ji;Jenny A. Harding;Richard Y. K. Fung

  • Affiliations:

  • Mechanical Engineering, National University of Singapore, Singapore 117576, Singapore;Industrial and Systems Engineering, The Hong Kong Polytechnic University, Hong Kong SAR, China;Industrial and Systems Engineering, The Hong Kong Polytechnic University, Hong Kong SAR, China;Wolfson School of Mechanical and Manufacturing Engineering, Loughborough University, United Kingdom;Department of Systems Engineering & Engineering Management, City University of Hong Kong, Hong Kong

  • Venue:
  • Computer-Aided Design
  • Year:
  • 2013

Quantified Score

Hi-index 0.00

Visualization

Abstract

Large amounts of online reviews, the valuable voice of the customer, benefit consumers and product designers. Identifying and analyzing helpful reviews efficiently and accurately to satisfy both current and potential customers' needs have become a critical challenge for market-driven product design. Existing evaluation methods only use the review voting ratios given by customers to measure helpfulness. Due to the issues such as viewpoints of interest, technical proficiency and domain knowledge involved, it may mislead designers in identifying those truly valuable and insightful opinions from designers' perspective. Thus, in this study, we initiate our work to explore a possible approach that bridges the opinions expressed by consumers and the understanding gathered by designers in terms of how helpful these opinions are. Our ultimate research focus is on how to automatically evaluate the helpfulness of an online review from a designer's viewpoint entirely based on the review content itself. We start our work by first conducting an exploratory study to understand the fundamental question of what makes an online customer review helpful from product designers' viewpoint. Through our study, we propose four categories of features that reflect designers' concerns in judging review helpfulness. Based on our experiments, it reveals that discrepancy does exist between both online customer voting and designers' rating. Furthermore, for the cases where review ratings are not steadily available, we have proposed to use regression to predict and interpret review helpfulness with the help of the aforementioned four categories of features that are entirely extracted from review content. Finally, using review data crawled from Amazon.com and real ratings given by design personnel, it demonstrates the effectiveness of our proposal and it also suggests that helpful product reviews can be identified from a designer's angle by automatically analyzing the review content. We argue that the study reported is able to improve designer's ability in business intelligence processing by offering more targeted customer opinions. It highlights the urgency to uncover sensible user requirements from such quality opinions in our future research.