A Framework on Early Decoupling Level Metric Assessment Based on NLP4RE

Modularity is recognized as a general and effective solution to tackle the complexity of system of systems in modern society. Despite the value and benefits of modularity, it has been a known challenge to measure and evaluate the modularity of complex systems in real life, especially early in the development life cycle such as in the requirement engineering phase. This paper proposes a framework that enables early assessment of the modular structure of a system based on system requirements. The framework combines (1) a Natural Language Processing for Requirements Engineering (NLP4RE) application developed together with the Systems Engineering Research Center (SERC), (2) a Decoupling Level metric, and (3) a novel Ablation procedure to quantify the decoupling and coupling effects of each design element in a system. We conduct a case study on an unmanned aircraft system (UAS) from SERC, for which the third and fourth authors of this paper have in-depth knowledge about. From this case study, we yielded interesting findings that align with expert’s intuitive understanding of the UAS. The case study indicates that our framework can provide early and quantitative assessment of the modular structure of a system based on the requirements. It has the potential to assist the architect in reasoning and designing a modular structure by reviewing the decoupling and coupling effects of system requirements and key terms.