NLP‐Enhanced Concern Mining for Viewpoint Pattern Extraction in Complex Systems

It is imperative to develop automated methodologies for decomposing complex system concerns into distinct prototype perspectives by transitioning from informal to rigorous, theoretical requirement engineering. In this paper, we propose an approach for automatic modeling and extracting stakeholder viewpoint patterns through mining concerns regardless of their functional or nonfunctional nature. This approach leverages NLP techniques to mine informal concerns, regardless of their functional or nonfunctional nature, and applies pattern recognition to concertize requirements across various similarity levels. This multidimensional separation of concerns facilitates early trade‐off analysis among overlapping and crosscutting concerns in terms of users, actions, and benefits within agile industrial systems. The applicability of the approach is analyzed on three industrial real‐world software systems, including the IoT‐based supply chain management system, Cloud ERP, and smart home online store. The system concerns are gathered through swarming GitHub requirement scenarios. Automatically extracted viewpoints for the systems are analyzed and then discussed regarding bounding effectiveness, precision, recall, scalability, and relevance determination after changes. The results show that the approach is promising to reach a certain level of maturity to facilitate requirement engineering in concern mining.