Towards Class Diagram Generation from User Stories Using LLMs

Integrating large language models into software modeling has unlocked new possibilities for automating laborintensive processes. One promising application is the automatic generation of Unified Modeling Language class diagrams from user stories, which traditionally requires significant manual effort and expertise. This paper takes the first step towards developing an application that utilizes LLMs to bridge artificial intelligence and software modeling. The proposed approach is assessed based on key metrics such as ease of integration and its ability to interpret ambiguous or incomplete requirements. Additionally, the paper examines the challenges of handling domain-specific constraints and the trade-offs between rule-based and fully automated methods, evaluating how they balance user control and system autonomy. By taking the initial steps towards streamlining class diagram generation and reducing reliance on manual modeling, this study contributes to advancing automation in software engineering.