Tracing the evolution of user requirements: a sentiment analysis -- based method for smart PSS innovative design

Smart product-service system (SPSS), as an outcome of the integrated development of information technologies and digital services, has become an essential outcome of the cyber-physical system (CPS) innovation. To support SPSS innovation design, it is meaningful to conduct requirement engineering with high efficiency and accuracy. However, the complex characteristics of SPSS and CPS present challenges in effectively utilising user-generated content, achieving accurate requirement analysis, and characterising the coupling relationship between product and service requirements. Existing approaches tend to be static and fail to address the dynamic nature of requirements for SPSS. To meet the challenges, this study proposes an integrated requirement elicitation, sentiment analysis, classification, importance calculation and mixed requirement evolution (ESCI-MR) method, which combines advanced deep learning algorithms and classic design theories. By eliciting requirements from online reviews on e-commerce platforms, natural language processing algorithms are employed to identify requirement keywords and sentiment information from extensive datasets. These keywords are then clustered using a clustering algorithm, and sentiment analysis is performed to enable requirement classification and importance computation. Following this, the method delineates the evolution of requirements by constructing multidimensional indicators. Applied to the sweeping robot system, the method's efficacy and advancements are validated, thereby bolstering the innovation process.