The Application of LLMs in the Analysis and Modeling of Software Requirements

In modern software engineering, requirements analysis and modeling are key steps in requirements engineering, influencing the subsequent system design and implementation. Traditional requirements analysis and modeling methods usually involve multiple stakeholders such as requirements providers and analysts working together collaboratively and iteratively, requiring a significant amount of manpower. It is of great significance to reduce the burden on requirements providers and analysts and improve the efficiency of analysis and modeling. In existing work, some merely utilize knowledge repositories to provide more knowledge in support of analysis or modeling, while others employ natural language processing (NLP) techniques to automate the analysis or modeling process. However, neither has relieved the burden on requirements providers and analysts. The emergence of large language models(LLMs) has brought new possibilities to requirements analysis and modeling. LLMs can combine knowledge base processing and understanding of large amounts of natural language data, providing a new automated requirement analysis method. This paper proposes an automated requirements analysis and modeling method for Support Vector Machine(SVM) classifier and LLMs - SVM-LLMs. The method is deployed based on the intelligent requirement service platform - “Wisdom Requirement Communication”, which combines SVM and LLMs to effectively improve the accuracy of requirements analysis and significantly reduce the time in the software development process.