Krishna Ronanki presenterar sin licentiatuppsats: Enhancing Requirements Engineering Practices Using Large Language Models
Large Language Models (LLMs) offer natural language interactions, technical insights, and task automation. However, integrating these models into Requirements Engineering (RE) for software development poses challenges. This thesis examines the technical capabilities and ethical considerations necessary for the facilitation of LLMs for RE, envisioning a systematic integration.
The studies that are part of this thesis use various methods including experiments, interviews, surveys, and a case study to evaluate LLMs’ performance in tasks like generating requirements, evaluating user stories, classifying requirements, and tracing dependencies. Findings show LLMs handle these tasks with different levels of success. Additionally, a review of ethical AI guidelines highlighted the constraints and challenges of using LLMs in RE.
Ten recommendations for using and integrating LLMs for RE in practice have been proposed, emphasising the importance of LLM’s role and user-friendliness, human oversight mechanisms and human-centric workflows to enhance their implementation effectiveness. Key conclusions underscore the need for a framework for task delegation between humans and LLMs, validating them in realistic environments, and considering human and organisational factors.