Abstract

The Web has changed the way that consumers express their opinions. They can now post reviews of products and express their opinions on almost anything on the websites. Potential customers often search online for product information and they often have access to hundreds of product reviews from other customers. Some of the reviews are found to be more helpful than other reviews as evidenced the potential customer’s helpfulness vote. This online word-of-mouth (WOM) behavior represents new and measurable sources of information.
Recent research has shown that helpfulness votes of customer reviews can have a positive influence on sales. While it is clear that helpfulness vote of a review is important, less is known about why certain pieces of online review are more helpful than others. Despite the fact that, customers encounter a variety of emotions in a purchase situation and those emotions are likely to be documented in the review, few researches have investigated how emotions elicited by the review affect the helpfulness of the review beyond the valence. Do discrete emotions have differential informational value in this case? Based on cognitive appraisal theory, in the first essay of this dissertation, I examine how specific emotions (hope, happiness, anxiety, disgust etc.) embedded in the review affect the helpfulness votes of potential customer. I adopt a quantitative content analysis (Latent Semantic Analysis) approach to measure emotions in these reviews.
In the second essay of the dissertation, I explore how the topics of online reviews differ between positive and negative reviews. Examination of real product reviews shows that there are thematic differences between them. Also, service related complains are found to be more helpful by potential customers. This enables us to better understand the conceptual differences in WOM .
Lastly, in the third essay, I compare two text mining techniques: Latent Semantic Analysis and Probabilistic Latent Semantic Analysis (PLSA) in extracting common themes in the positive and negative product reviews. Results shows that the choice of text mining approaches should be based on the goal of the marketing researcher. If the goal is to learn about a specific brand, PLSA might reveal more specific information. However, if the goal is to learn about important aspects of a broader product category, LSA works better in terms of interpretability.