Artificial intelligence for generation and verification of UML and BPMN diagrams

Software modelling has grown in importance since the end of the 20th century, and its importance continues to grow with IT systems becoming increasingly important, complex, and larger in scale. This dissertation addresses the intersection between artificial intelligence (AI) and software modelling, focusing on the generation of Unified Modeling Language (UML) and Business Process Model and Notation (BPMN) diagrams by AI. Being that UML plays a key role in representing complex and large-scale information systems structurally and BPMN helps communicate effectively business processes, this research is focused on the potential capability of AI, specifically GPT 3.5 and GPT 4.0, to generate and create these diagrams automatically from descriptive problem statements. By creating a scoring system to evaluate diagrams and syntax errors in the solutions generated by the three actors (human, GPT 3.5 and GPT 4.0) as well as using statistical analyses such as t-tests and ANOVA, the research aims to assess whether AI-generated solutions can achieve or exceed the efficiency of human-generated solutions in terms of accuracy and effectiveness. The final statistical results suggest that GPT 3.5 and GPT 4.0 do not outperform human performance in generating BPMN diagrams. This outcome highlights the difficulties and limitations of GPT(Generative Pre-trained models) and Language Large Models in apprehending complex contexts and requirements, generating graphic representations, and possessing limited domain-specific knowledge. It also underscores the importance of ongoing exploration and technological progress to enhance AI capabilities in this field.