Semi-Automated Mining of Customer Reviews to Identify Design Problems of Cantilever-Style Bassinets

Abstract Customer reviews posted online provide a massive dataset that can identify product defects. Locating informative reviews of a specific product is straightforward. However, it can be challenging to extract information that informs product designers of potential defects and drives useful testing to investigate the potential defect. Such data extraction can be manually performed in a reliable and effective manner by skilled engineers. However, such an approach can be time-consuming and expensive. On the other hand, completely automated data extraction would likely yield generic results that are not particularly useful. This paper describes a semi-automated process wherein researchers develop an informed search algorithm that is based on an initial engineering assessment of a product that identifies potential design problems and words sets that are associated with the potential defects. An automated search process then operates on the review database to greatly streamline the data analysis process. The product reviews that indicate product defects are grouped together and associated with design features. The design features that are linked to the largest classes of complaints are then rigorously tested to evaluate the credibility and accuracy of the online complaints. The process is illustrated via a case study and testing of a cantilever-style bassinet.