Improving Requirements Engineering for Space Systems using Large Language Modeling

Requirements engineering (RE) is a critical phase in system engineering for space systems, in which mission needs are translated into specifications that are verifiable and traceable to guide design, development, and verification. The efficiency of this process is measured using a metric that leads to the quantification of accuracy, timeliness, completeness, and the efficiency of resource utilization, while reducing the risk of errors and the likelihood of reuse. The task of Concept Recognition (CR) as defined in the field of Natural Language Processing (NLP) is useful for forming, applying, and identifying abstract concepts, such as categorizing objects, attributes, or relationships in linguistic contexts, thereby making the development of requirements more efficient.Achieving operational efficiency in space systems engineering requires the reuse of knowledge from archives of past system data. However, very little of this is actually labelled which presents an opportunity for both the transformer architecture and transfer learning to be applied in a manner similar to those that have successfully transformed the field of NLP. Earlier research applied transfer learning via further pretraining (FPT) and fine-tuning (FT) to pre-2021 versions of the Bidirectional Encoder Representations from Transformers (BERT) family of models, whereas this research extends that work to more recent, larger BERT models. Lastly, given the substantial advances in recent frontier AI models, a prompt engineering library has been developed for the CR task and an agentic workflow is designed to iteratively optimize the prompt, thereby automating the prompt engineering process.

Author

• Fargnoli, Joseph Domenic

Language

• en

Keyword

• DeBERTa
• Space Systems Engineering
• Concept Recognition
• Requirements Engineering
• Natural Language Processing
• Large Language Models

Date created

• 2026

Type of Work

• Dissertation

License

• All rights reserved

Rights statement

• In Copyright

GW Unit

• Computer Science

Degree

• D.Engr.

Advisor

• Hassan, Walid
• Etemadi, Amir

Committee Member(s)

• Veit, Daniel

Persistent URL

•  https://scholarspace.library.gwu.edu/etd/2v23vv414