PROFES 2024
Mon 2 - Wed 4 December 2024 Tartu, Estonia

Software and systems development projects in regulated domains need to provide evidence on their compliance to standards. Such standards often come as comprehensive documentation, i.e., documents, which need to be tailored to and interpreted for a particular project. Utilizing such documentations, it is desirable for companies working in regulated domains to provide tool support with regard to measurement systems that, eventually, help determine the product’s quality and to support the compliance analysis. In this paper, we present an AI-supported approach to use a standard’s documentation for generating artifact models. Such models are generated in a machine-readable format and, therefore, help companies create measurable items to be included in their metrication and measurement systems. Using selected ECSS standards from the European Space Agency as cases, we present our approach, the prompt engineering for the extraction of artifacts, and we illustrate the opportunities to generate comprehensive artifact models. Our findings show that, given sufficient information is available, the generation of artifact models is possible to a large degree with an average completeness of 99.64% and an average precision of 67.52% thus laying the foundation for building more efficient measurement systems.