Assessing User and Manufacturer Perceptions of Fitness Trackers through Amazon Review Analysis

Abstract

Fitness trackers encourage people to be more active, people with obesity track their diet, and older adults to understand their health by knowing their heart rate. Companies display advertisements for these types of products and describe them as beneficial. However, users are looking for the products that best suit their personal needs, for which they often review the opinions of other users on e-commerce platforms such as Amazon. In this research, we study the opinion of users who have used physical activity trackers, to assess whether their satisfaction meets the quality offered by the manufacturer. Through: the use of natural language processing techniques, and the analysis of information provided in Amazon reviews. A sentiment analysis was carried out based on the technical characteristics of the device. We employ transfer learning of a Transformer-based language model (RoBERTa: Robustly optimized BERT Pretraining Approach). Which was retrained for two classification problems in independent modules, the first module classified 20 technical aspects of the device (93% precision), and the second module classified user sentiments (70% precision). A comparison was made between the average feeling of the user vs. the manufacturer, getting a 3.11 for users and 3.99 for manufacturers with a difference of 0.88, concluding that the user’s sentiment having a lower expectation than the manufacturers.