User stories capture requirements in agile development in semi-structured natural language, but their analysis becomes complex as requirements evolve. Knowledge Graphs (KGs) have been proven useful to partially automate this analysis. Existing works automate user story modeling via Natural Language Processing (NLP), but have limited precision and come with complex implementations. A recent approach by Arulmohan et al. AMM23 employs Large Language Models (LLMs); however, it relies on specific model providers and lacks a knowledge graph creation. This thesis introduces a novel, end-to-end methodology for automated KG generation from user stories, centered on the UserStoryGraphTransformer (USGT), a LangChain-based module that provides a provider-agnostic, LLM-based modeling component that can be easily (re)evaluated against a given ground truth, delivering a flexible solution for modern requirements engineering
Thayná Camargo da Silva (Thu,) studied this question.