Date of Award
Doctor of Philosophy
Huimin HZ Zhao
Mark MS Srite, Yang YW Wang, Abhijeet AG Ghoshal
Econometrics, Helpfulness, Image analysis, Online reviews, Product returns, Text mining
Since the early-2010s, with the technology development, user-generated pictures in online reviews have flourished. As large volumes of user-generated pictures are being generated along with online reviews, online businesses want to learn whether the pictures can improve the information quality (helpfulness) of reviews and what features of the pictures would be meaningful for consumers in purchasing a product online. Nevertheless, in the academic area, little is known about the effects of visual information, i.e., a review picture, while text-mined features have been widely investigated. In this dissertation research, we strive to tackle the economical role of user-generated pictures posted along with online reviews and to see how the helpfulness of review pictures is perceived differently across national cultures.Essay 1 explores how visual features in an online review posted along with review texts improve review helpfulness and reduce product returns. We mainly extract a feature of situational cues from review pictures based on two types of information: environmental cues for product usage and human activity-driven aspects. As counterparts of the visual information, we text-mine a functionality feature as well. To control for other visual impacts on review helpfulness and product returns, we additionally image-mine such attributes as a malfunction of a product, image saturation, clarity, and sharpness. The main findings are that the visual feature, which better illustrates product functionality, is perceived as being helpful by consumers and helps decrease product returns by assisting consumers in making an informed decision. Furthermore, we found a picture superiority effect, that is, shoppers prefer to use image context rather than reading review texts when they have higher information processing costs and a helpful review with a picture helps reduce product returns as well whereas one without does not. Essay 2 broadens our perspective of online reviews beyond the local e-commerce market. In online review systems, consumers vote for a review fundamentally because they trust the review information and/or reviewer and thus are willing to rely on the given information. Yet, consumers may build trust based on trust antecedents differentially across national cultures–some consumers may depend more on the information quality of a review whereas others rest on the source (reviewer) credibility–according to the elaboration likelihood model theory. Thus, this essay takes a cross-cultural look at online reviews. In this research, both image and text mining techniques are employed. We mainly extract a human appearance feature from review pictures (i.e., information quality) and profile photos (i.e., source credibility) for comparison as counterparts. From review texts, we derive topic diversity (i.e., information quality) using topic modeling and measure review sentiment through sentiment analysis. We found that the perception of review helpfulness tends to be discerned across national cultures. More interestingly, the same content in a different layout, here human appearance in a profile photo versus in a review picture, is differentially helpful for consumers across national cultures. This study can benefit online businesses in terms of global review system management.
Kim, Youngeui, "Analysis on User-Generated Pictures in Online Reviews" (2023). Theses and Dissertations. 3172.
Available for download on Thursday, May 22, 2025