Topic modelling-based decision framework for analysing digital voice of the customer

Abstract

The aim of this study is to introduce a decision framework that integrates Quality Function Deployment (QFD) methodology and topic modelling to facilitate analysing online reviews, that is, a digital source of voice of the customer (VoC), to extract the true customer needs for developing a product/service. Within the framework, after collecting and preprocessing customer reviews, Latent Dirichlet Allocation is used for grouping them as topics. Most representative reviews are determined based on corresponding topic distributions, and QFD methodology is employed to clarify these reviews and reveal true customer needs. For demonstration, this framework was used to analyse 83,545 reviews on Italian restaurants gathered from Yelp.com via topic modelling with a more objective perspective. The number of documents to be examined was decreased by approximately 95%. Thereupon the completion time of customer voice table, one of the most important tools in QFD studies, reduced dramatically. Besides, text mining process is enhanced for retrieving meaningful information from customer reviews by using QFD tools and their tailored approach. This study is the first attempt that integrates QFD and topic modelling in order to increase the efficiency of the VoC clarification process.