Language is the Bridge

2025-05-26 · Source: Intentional Arrangement · Field: Technology & Digital — Artificial Intelligence & Machine Learning, Data Science & Analytics, Software Development & Engineering · Depth: Expert, extended

Summary

The article emphasizes that language is the fundamental "substrate and bridge" for both Large Language Models (LLMs) and ontologies, asserting that stripping away linguistic scaffolding renders these systems unintelligible and unactionable. It references the 2020 "octopus paper" by Emily Bender and Alexander Koller, which illustrates how systems trained only on linguistic form produce "fluent nonsense" without real-world grounding. The author extends this concept to ontologies, arguing that an internally valid OWL ontology is "externally inert" if its terms are not bound to controlled vocabularies or taxonomies reflecting actual organizational usage. The piece highlights that labels are crucial for human and machine interoperability, citing Wikidata's enforcement of unique label-description combinations and the OOPS! pitfall scanner's identification of "missing annotations" as a common issue. Ultimately, the "Ontology Pipeline™" is presented as an engineering framework for building shared linguistic agreement across controlled vocabularies, taxonomies, thesauri, ontologies, and knowledge graphs to bridge the "knowing-doing gap" in organizations.

Key takeaway

For executives overseeing data strategy and knowledge management, recognize that the "knowing-doing gap" in your organization is fundamentally a language gap. Prioritize investment in the "Ontology Pipeline™" by establishing controlled vocabularies, taxonomies, and thesauri before formalizing ontologies or knowledge graphs. This foundational work ensures shared understanding, enabling coordinated action and preventing costly decision paralysis stemming from ambiguous terminology.

Key insights

Shared language and explicit linguistic scaffolding are essential for making LLMs and ontologies actionable and trustworthy.

Principles

• Form is not meaning; systems need grounding in real-world context.
• Labels are the human interoperability layer for semantic systems.
• Vocabulary control is a precondition for semantic operations.

Method

The Ontology Pipeline™ systematically builds shared linguistic agreement through layered structures: controlled vocabularies, taxonomies, thesauri, ontologies, and knowledge graphs, ensuring terms are bound to human practice.

In practice

• Integrate retrieval augmentation and tool use for LLM grounding.
• Enforce unique label-description combinations in knowledge graphs.
• Prioritize "alt labels" to resolve vocabulary mismatches.

Topics

• Language Models
• Ontologies
• Knowledge Graphs
• Semantic Interoperability
• Ontology Pipeline

Code references

• edmcouncil/fibo

Best for: Executive, AI Architect, Director of AI/ML, Consultant

Related on AIssential

• A Practitioner's Guide To Taxonomies, Part I — Taxonomies offer high ROI as low-fidelity semantic structures, enabling context and inference without full ontological complexity.
• Ch 11 - Context, Intent, and Action: The Semantic Foundation — Explicitly capturing shared meaning through semantics is crucial for effective data modeling, especially with AI integration.
• The Ontology Pipeline™, Refresh — Effective semantic knowledge infrastructure requires a structured pipeline, human expertise, and robust governance, not just AI-generated outputs.
• What Makes Knowledge Graphs AI-Ready — Knowledge Graphs for AI bridge siloed data semantics with organizational ontology for machine-readable reasoning.
• My upcoming book 🤸‍♂️ — The "Ontology Pipeline" framework systematically builds semantic knowledge infrastructures from foundational vocabularies to knowledge graphs.
• Knowledge Engineering for Search and Content: A Practical Guide — Structured understanding of content and user intent is crucial for effective LLM-powered search and content experiences.

Open in AIssential →

Editorial summary, takeaway, and curation by AIssential. Original article published by Intentional Arrangement.