A Retrieval-Augmented NLP Framework for KDC Research Knowledge
Over the past decades, the Knowledge and Development Centre Mainport Schiphol has built an extensive archive of research reports and innovation studies in Air Traffic Management (ATM). While this represents significant intellectual capital, the scale and diversity of documents make systematic knowledge reuse increasingly challenging. This project develops a Retrieval-Augmented Generation (RAG) framework tailored to KDC’s research environment. Documents are segmented and stored in a dedicated vector embedding database, enabling semantic similarity search instead of keyword-based retrieval. Retrieved fragments are supplied to a Large Language Model (LLM), allowing it to generate structured, context-aware responses grounded in KDC’s own research outputs. The challenge is not implementing RAG itself, but ensuring reliable reasoning and controlled generation in a research-critical context. The system must handle domain-specific terminology, maintain traceability to source material, and minimise hallucinations.
Beyond improving knowledge access within KDC, this work lays the foundation for applying RAG-based, context-aware generation within LVNL, supporting future knowledge management and research alignment across the Dutch ATM ecosystem.