A Study of Online Review Helpfulness

Google Drive Link

This study aims to explain why some reviews are perceived as being more helpful than others and, in this manner, suggest some effective e-WOM management strategies for different product types. Retail websites might use these in developing product review writing guidelines that may foster more effective communication systems. This way, these online social interactions could be stimulated, which benefits the customer by reducing their purchasing uncertainty and the online retailer by increasing customer “stickiness” to provided services. Besides this, we aim to develop a machine learning algorithm that retailers could use in initially assessing the helpfulness of product reviews. This will considerably facilitate a better organization of reviews that will weaken the established Matthew effect, which implies that top reviews get more helpful votes due to their position within the review page compared to lower ones (Wan, 2015). Hence, this way, the customer burden of analyzing the ever-increasing enormous amount of review data will be diminished, leading towards a helpfulness assessment system that is not influenced anymore by the three previously identified types of bias: imbalanced vote bias, winner circle bias, and early-bird bias (Liu et al., 2007). In the rest of the research paper, we first discuss the previous studies conducted within the WOM and e-WOM research fields. Then the research models are developed, around which hypotheses that are going to be tested are formulated. After this, we discuss in detail the data this study makes use of. Subsequently, we also discuss the methods that are implemented in analyzing our data. Then, we present the result of the analysis and the performance of the built-in models. And finally, we discuss the effects and initiate a discussion around those by summarizing our conclusions and limitations.