Towards Automated or Assisted Requirements Extraction and Analysis Using an LLM-Based Multi-Agent System: A Case Study

The increasing variety and complexity of products presents companies with ever greater challenges when selecting and engineering suitable components and systems. A pivotal step in this process is the collection and documentation of requirements, which is conducted manually, necessitating substantial resources, particularly when the underlying data is in unstructured formats. In this publication, we propose an innovative approach to address this issue. Specifically, we present a multi-agent system that is developed based on large language models. This system facilitates the automated or at least assisted analysis of requirements extracted from unstructured data, with a particular emphasis on the processing of text and image data. Initial results show that the approach presented delivers promising results.