Semi-Automated Knowledge Engineering and Process Mapping for Total Airport Management

A team of researchers has developed a novel AI framework to tackle the notoriously complex challenge of documenting and managing airport operations. The system, detailed in a new arXiv paper, addresses the data silos and semantic inconsistencies that plague the aviation industry by constructing a domain-grounded, machine-readable Knowledge Graph (KG). It does this through a dual-stage fusion of traditional symbolic Knowledge Engineering (KE) and modern generative Large Language Models (LLMs). In this scaffolded approach, expert-curated KE structures guide LLM prompts to discover semantically aligned knowledge triples from vast, unstructured textual corpora, effectively automating the synthesis of complex operational workflows.

A key technical finding challenges conventional AI wisdom. The team evaluated their methodology using the Google LangExtract library and discovered that, contrary to prior observations of long-context degradation in models like GPT-4 or Claude, document-level LLM processing actually improves the recovery of non-linear procedural dependencies compared to localized segment-based inference. This is critical for mapping interconnected airport processes. Most importantly, the framework bridges the gap between generative AI's "black-box" nature and the absolute transparency required for safety-critical operations. It fuses a probabilistic model for discovery with a deterministic algorithm that anchors every single data extraction to its source document, guaranteeing 100% traceability and verifiability for airport managers and regulators.