Bachelor Thesis: LLM4Software Modeling

The goal of this thesis is to evaluate the suitability of Large Language Models (LLM) for software modeling using the UML. Thereby, it shall be investigated whether LLMs are capable of correctly generating the most common UML diagram types (use case, class, sequence, and state machine) from literal specifications. In the course of this, advanced LLM tools like Prompt Engineering [1], Language Model Programming [2], and Grammar Prompting [3] shall further be evaluated w.r.t. their potential in improving the retrieved results. The thesis concludes with a dedicated proposal (based on the evaluated tools) for enabling software modeling using LLMs.

Student: 1

Contact:
Philipp.Zech@uibk.ac.at

[1] https://arxiv.org/abs/2302.11382
[2] https://lmql.ai
[3] https://neurips.cc/virtual/2023/poster/72512