Abstract
Online reviews published on the e-commerce platform provide a new source of information for designers to develop new products. Past research on new product development (NPD) using user-generated textual data commonly focused solely on extracting and identifying product features to be improved. However, the competitive analysis of product features and more specific improvement strategies have not been explored deeply. This study fully uses the rich semantic attributes of online review texts and proposes a novel online review–driven modeling framework. This new approach can extract fine-grained product features; calculate their importance, performance, and competitiveness; and build a competitiveness network for each feature. As a result, decision making is assisted, and specific product improvement strategies are developed for NPD beyond existing modeling approaches in this domain. Specifically, online reviews are first classified into redesign- and innovation-related themes using a multiple embedding model, and the redesign and innovation product features can be extracted accordingly using a mutual information multilevel feature extraction method. Moreover, the importance and performance of features are calculated, and the competitiveness and competitiveness network of features are obtained through a personalized unidirectional bipartite graph algorithm. Finally, the importance performance competitiveness analysis plot is constructed, and the product improvement strategy is developed via a multistage combined search algorithm. Case studies and comparative experiments show the effectiveness of the proposed method and provide novel business insights for stakeholders, such as product providers, managers, and designers.
| Original language | English |
|---|---|
| Journal | INFORMS Journal on Computing |
| Early online date | 17 Jan 2024 |
| DOIs | |
| Publication status | E-pub ahead of print - 17 Jan 2024 |
Bibliographical note
Acknowledgments:The authors thank the area editor and the anonymous reviewers for their constructive comments. This work was supported by the National Natural Science Foundation of China (NSFC) under Project 72071151, and the Natural Science Foundation of Hubei Province, China (2023CFB712).