Well-Formed Quality of System Requirements for Verifying to ISO 29148-2018: A Natural Language Processing (NLP) Based Framework and Quantitative Metric

Abstract The system requirements directly affect the quality, cost, and performance of system architecture developed in the detailed design phase, necessitating a comprehensive approach to ensure accuracy, completeness, and clarity. A well-defined set of requirements lays the foundation for developing effective and efficient systems, guiding the architectural design process toward meeting stakeholder expectations while optimizing resource utilization and minimizing potential risks associated with ambiguous or poorly specified requirements. Methodologies for assessing the quality of individual requirements generated during the initial conceptual system design phase are crucial. In this paper, we outline a rule-based natural language processing (NLP) framework for assessing the well-formed quality of individual requirements that conforms to the ISO/IEC/IEEE 29148:2018 (E) requirement standard. We leverage advanced linguistic analysis to automatically evaluate each of the nine essential characteristics — necessary, appropriate, unambiguous, complete, singular, feasible, verifiable, correct, and conforming — prescribed by ISO/IEC/IEEE 29148:2018 (E) to ensure the clarity, completeness, and consistency of individual requirements. The framework assesses requirements at various levels, including token, span, syntactic structure, and sentence levels, to determine compliance with the defined characteristics of well-formed requirements. Furthermore, this work introduces a quantitative metric, the well-formed requirement quality index, which amalgamates these nine characteristics into a scalar value within the range of [0, 1], thereby facilitating a comprehensive assessment of requirement quality. The efficacy of this framework is demonstrated through the evaluation of select requirements samples, offering a means to assess requirement quality and perform quantitative comparisons of individual requirements’ qualities.