Abstract:

Modern systems have become increasingly complex, and traditional engineering methods often struggle to handle these complex problems. Model-based systems engineering (MBSE) could rapidly and accurately grasp system behavior and performance by constructing system models, analyzing and simulating systems, thus effectively addressing complexity problems. The systems modeling language (SysML) can aid in clear understanding and description of systems. Previously, natural language-based system description methods were prone to misunderstandings and inaccurate expression issues. Therefore, SysML diagram automatic generation based on NLP technology is currently a research area with great academic value, but current research in this area is relatively limited. This study proposed a method named natural language-driven use case diagram automatic generation, abbreviated as NLUCD. Firstly, it employed pre-determined language rules to preprocess input text. Then it applied NLP tools for text segmentation, stemming, and lemmatization. Next, it utilized a fine-tuned BERT (bidirectional encoder representations from transformers) model for named entity recognition and syntax dependency analysis. Subsequently, it defined three sets of rules to extract participants, use cases, and relationships in the use case diagram, corresponding to the elements and relationships in the use case diagram. Finally, the drawing tool completed the generation of the SysML use case diagram. Although the proposed method is currently mainly applicable to English text, it provides new ideas and perspectives for the automation and intelligence of the system design field, having both theoretical and practical value.

Key words: MBSE, SysML use case diagram, NLP, custom rules, BERT