Towards Lean Automation: Fine-Grained sentiment analysis for customer value identification. (July 2022)

• Record Type: Journal Article
• Authors: Xiao, Yan; Li, Congdong; Thürer, Matthias; Liu, Yide; Qu, Ting
• Is Part Of: Computers & industrial engineering. Volume 169(2022)
• Publication Date: 2022-07
• Subjects: Product Development -- Fine-grained Sentiment Analysis -- Deep Learning -- Case Study; Engineering -- Data processing -- Periodicals; Industrial engineering -- Periodicals; 620.00285
• DOI: 10.1016/j.cie.2022.108186
• Languages: English
• ISSNs: 0360-8352

Abstract

Highlights: A new approach for fine-grained sentiment analysis is presented. Integrates pre-training language, conditional random field, and linguistic knowledge model. The new approach outperforms six common alternative approaches. Ablation experiments show that our approach is parsimonious; all components are needed. A use case exemplifies how our approach can be used to guide process improvement. Abstract: That customer value should drive product development and production is a basic tenet of the Toyota Production System and Lean. Traditional means to extract what the customer wants often focus on customer surveys. But surveys are time consuming and costly. At the same time, there exists a large amount of customer comments in online reviews that is easily accessible, whilst the advances of data science, for example as part of Lean Automation, provide new means to extract information from this data. In this context, a new approach to fine-grained sentiment analysis of Chinese consumer data is developed. The new approach integrates pre-training language model, conditional random field model and linguistic knowledge model. The new approach is shown to outperform traditional approaches in a comparison experiment, while an ablation experiment shows that our new approach is parsimonious, i.e., all three constituting components are needed. Finally, a use case is presented that exemplifies how our new approach can support managers in identifying customer value (through positive evaluations), and most importantly guide Lean improvement, through detailed information on characteristics that are evaluated negatively, ranked according to customer importance. Findings have important implications for research and practice.