Natural Language Processing for Effective Use of User Stories: Systematic Literature Review

User stories are an essential element of agile software development due to their flexible structure and simplicity. The diverse range of data sources and formats utilized for documenting requirements poses challenges for development teams, who often have to translate client needs into user stories manually, a process that is both time-consuming and error-prone. To address this, interest has grown in automating user story generation. A systematic literature review (SLR) was conducted to assess current methodologies, focusing on the types of corpora used, the natural language processing (NLP) techniques employed and the methods used to evaluate the quality of generated user stories. Our review reveals a significant shortage of publicly available corpora, which limits progress in this field, particularly for machine learning-based approaches. It also identifies a wide range of techniques and highlights the lack of standardized methods for evaluating user story quality. More recent developments involving Large Language Models (LLMs) offer new opportunities for both user story generation and quality evaluation, but also raise concerns regarding hallucination, consistency, semantic fidelity, reproducibility, and the need for human validation. Overall, the findings highlight the need for publicly available datasets, standardized evaluation frameworks, and rigorous empirical studies to assess the effectiveness and reliability of automated and LLM-based user story generation approaches.