Automatic Generation of Ontology for Aircraft Maintenance Applications

Kurzfassung

This paper presents an automated pipeline for transforming aircraft maintenance documentation into semantically rich ontologies and knowledge graphs to support nextgeneration defect management. Leveraging S1000D schema as a semantic backbone, the system integrates layout-aware text processing, BERT-based entity and relation extraction, and synthetic data augmentation to enhance precision and recall. A multi-stage novelty filter ensures lexical, semantic, and logical consistency before ontology population and reasoning with SWRL rules. Comparative evaluation against LLM-only extraction demonstrates improved accuracy and explainability. The resulting ontology-driven knowledge graph forms a foundational resource for Retrieval-Augmented Generation (RAG) and AI-assisted MRO decision-making.