Impact Analysis of using Natural Language Processing and Large Language Model on Automated Correction of Systems Engineering Requirements
Lucas Henrique Marchiori, Arthur Hendricks M. de Oliveira, Pedro Almeida Reis, Fernando Sarracini JĂșnior, Mairon Sena Cavalcante, Jonathan Vinicius de Lima, Luis Fernando Soares (Ford Motor Company)
Keywords
Requirements;Systems Engineering;Large Language Models;Natural Language Processing
Abstract
The increasing complexity of Electronic Control Units (ECUs) in the Automotive Industry due to the integration of more sophisticated vehicle features led to a greater need for robust Systems Engineering (SE) to define and implement efficient solutions. In this context, requirements emerge as a critical part of the communication between cross-functional teams. The more complex systems become, the more requirements are needed to define them. Misalignment, lack of information and ambiguity on requirements impact the entire development process, resulting in issues later, harder to be fixed. Some studies are being applied to evaluate techniques using Natural Language Processing (NLP) and how it can replace extensive peer reviews, identifying weaknesses in requirements earlier in the process, avoiding wasted time and large financial losses. Normally, NLP is combined with templates such as Easy Approach Requirements to Syntax (EARS), or other techniques based on rules like the INCOSE rules to define metrics and evaluate the quality of requirements in automated way. The focus of this study is to enhance the requirements evaluation algorithm by combining NLP with Large Language Models (LLMs) and adding the ability to provide corrected requirements to Systems Engineers.