TY - GEN
T1 - Large Language Model-Assisted Natural-Language-To-Sql Pipeline for Project Management in Aviation Certification
AU - Venda, Ana Margarida Henriques
AU - Costa, Ruben Duarte Dias da
AU - Rega, Bruno Do Monte Costa
AU - Jardim-Goncalves, Ricardo
N1 - info:eu-repo/grantAgreement/FCT/Avaliação UID 2023%2F2024/UID%2F00066%2F2025/PT#
The authors acknowledge the European Commission for the support and funding under the scope of Horizon Europe MaaSAI Project (Grant Agreement Number 101177368).
Publisher Copyright:
© 2026 IEEE.
PY - 2026
Y1 - 2026
N2 - Standard reporting tools and information systems in project management often lack the flexibility required to adapt to evolving project scopes without technical intervention. Across various industries, this rigidity creates severe bottlenecks, delaying critical decision-making for Project Managers (PMs). This operational latency becomes particularly detrimental in safety-critical environments, where continuous oversight and high precision are demanded. To address this limitation, this work explores a Natural Language-to-SQL (NL2SQL) pipeline designed to translate natural language inquiries into executable database queries. Unlike generalized end-to-end AI solutions, the proposed architecture strictly constrains the Large Language Model to a semantic parsing role, enforcing execution through a controlled backend. Using anonymized industrial datasets, this study evaluates the impact of this decoupled approach on operational agility, technical dependency reduction, and adherence to the strict data sovereignty requirements of the aviation domain.
AB - Standard reporting tools and information systems in project management often lack the flexibility required to adapt to evolving project scopes without technical intervention. Across various industries, this rigidity creates severe bottlenecks, delaying critical decision-making for Project Managers (PMs). This operational latency becomes particularly detrimental in safety-critical environments, where continuous oversight and high precision are demanded. To address this limitation, this work explores a Natural Language-to-SQL (NL2SQL) pipeline designed to translate natural language inquiries into executable database queries. Unlike generalized end-to-end AI solutions, the proposed architecture strictly constrains the Large Language Model to a semantic parsing role, enforcing execution through a controlled backend. Using anonymized industrial datasets, this study evaluates the impact of this decoupled approach on operational agility, technical dependency reduction, and adherence to the strict data sovereignty requirements of the aviation domain.
KW - Large Language Models (LLM)
KW - Natural Language-to-SQL (NL2SQL)
KW - Project Management
KW - Safety-Critical Systems
KW - Semantic Parsing
UR - https://www.scopus.com/pages/publications/105047039225
U2 - 10.1109/YEF-ECE70590.2026.11614656
DO - 10.1109/YEF-ECE70590.2026.11614656
M3 - Conference contribution
AN - SCOPUS:105047039225
T3 - Proceedings - 2026 10th International Young Engineers Forum on Electrical and Computer Engineering, YEF-ECE 2026
SP - 33
EP - 38
BT - Proceedings - 2026 10th International Young Engineers Forum on Electrical and Computer Engineering, YEF-ECE 2026
PB - Institute of Electrical and Electronics Engineers (IEEE)
T2 - 10th International Young Engineers Forum on Electrical and Computer Engineering, YEF-ECE 2026
Y2 - 19 June 2026
ER -