Practical Research on Intelligent Requirements Analysis and Test Case Generation Driven by Large Language Models

The increasing complexity of software systems underscores that the quality and efficiency of requirements engineering, as the initial phase of the software lifecycle, directly impact project outcomes. Traditional requirements analysis heavily relies on manual effort, facing challenges such as low efficiency and difficulty in maintaining consistency. The rapid development of generative artificial intelligence, particularly large language models (LLMs), offers new tools to address these challenges. This study explores the practical application of LLMs in the software requirements analysis phase, focusing on the automated extraction and structuring of requirement elements, as well as the assisted generation of preliminary test cases. This paper proposes a “five-step” human-AI collaborative workflow incorporating large models and elaborates on its implementation steps through a retrospective case study of an intelligent warehouse management system of an intelligent warehouse management system. Practice demonstrates that this method can partially free requirements analysts from repetitive information extraction and formatting tasks, enhance the structured nature of requirements documentation, and provide valuable initial input for downstream test design. Although limitations such as model “hallucination” exist and necessitate strict manual review, this research confirms the positive potential of LLMs in standardizing and improving the efficiency of requirements engineering within software factories. Finally, the paper summarizes key experiences and challenges from the practice, providing a reference for peers seeking to introduce AI assistance in similar scenarios.