Software requirements analysis with large language model based artificial intelligence aid

Safety-critical systems (SCSs) are systems where failure can lead to catastrophic consequences, including loss of life, environmental damage, and significant economic loss. In such contexts, requirements analysis is a foundational phase that significantly affects the safety, reliability, and verifiability of the system. However, traditional practices for analyzing requirements often struggle with inefficiencies, ambiguity, and human error. With the advent of advanced artificial intelligence (AI) technologies—especially natural language processing (NLP), machine learning (ML), and large language models (LLMs)—novel approaches are emerging to automate and enhance the precision and scalability of requirements analysis. This work proposes a novel software requirements analysis method, which utilizes a real-time service system that enables asking questions and receiving answers about PDF documents. The service system involves document loading, chunking, embedding, storage in a vector database, and retrieval using semantic similarity. The final response is generated using an LLM. An experimental study has been performed to demonstrate the feasibility and effectiveness of the proposed method.