Open Ontologies: Tool-Augmented Ontology Engineering with Stable Matching Alignment

Open Ontologies is an open-source Rust-based system that combines LLM-driven ontology engineering with formal OWL reasoning and stable matching alignment. The research demonstrates that stable 1-to-1 matching is the critical factor for ontology alignment quality, achieving F1 scores competitive with state-of-the-art systems, while structured tool access via Model Context Protocol significantly outperforms raw file reading for LLM interaction.

Open Ontologies addresses a fundamental challenge in ontology engineering by systematically isolating what actually drives alignment quality. The research reveals that stable matching algorithms dominate performance gains—removing this component causes F1 scores to drop from 0.832 to 0.728 on the Anatomy track, while varying signal weights produces negligible differences. This finding challenges conventional wisdom about feature engineering complexity and suggests that algorithmic elegance often trumps parameter tuning.

The most provocative discovery concerns how LLMs interact with structured data. When presented a raw OWL file, the same model performs substantially worse (F1 0.323) than with no file context at all (F1 0.431), yet achieves F1 0.717 through structured Model Context Protocol tool access. This reveals a critical insight: LLMs struggle with unstructured syntax and benefit dramatically from programmatic abstraction layers. The finding has implications beyond ontology engineering—it suggests that AI systems require purpose-built interfaces rather than direct access to raw technical formats.

The release as a single Rust binary under MIT license positions this work as immediately deployable infrastructure rather than academic exercise. For developers and enterprises managing knowledge graphs or semantic systems, the ablation results provide actionable guidance: prioritize stable matching implementation over complex feature weighting. The stark performance gap between raw file access and tool-augmented interaction suggests organizations should architect AI-facing systems around structured APIs, not document dumps. These findings bridge theoretical computer science with practical system design, offering testable principles for production-grade ontology systems.

• →Stable 1-to-1 matching is the dominant factor in ontology alignment, with F1=0.832 on Anatomy track, exceeding precision of competing systems
• →Signal weight variations produce negligible F1 differences (<0.004), indicating that algorithmic structure matters far more than parameter tuning
• →LLMs perform dramatically better with structured tool interfaces (F1=0.717) than raw file access (F1=0.323), revealing a critical design principle for AI systems
• →The system achieves competitive performance on the Anatomy track while performing more modestly on the Conference track (F1=0.438), suggesting domain-specific variation
• →Open-source MIT-licensed release as single binary enables immediate adoption in production ontology engineering workflows