Explicit Semantics for AI Applications: Ontologies in Practice

Dr. Jesús Barrasa presents on the relevance of ontologies in modern AI applications. He highlights the tendency of AI systems to struggle with implicit and fragmented semantics, emphasizing that explicit semantics through ontologies and knowledge graphs can transform the design and operation of AI systems. The session is divided into key sections covering the basics of ontologies, practical applications, and methods for building and evaluating them.

Main Sections

• Introduction to Ontologies: Ontologies are essential for defining explicit semantics in AI. Barrasa discusses key components of ontologies, including classes, properties, and relationships, as well as their application in grounding AI systems and enabling reasoning.
• Practical Application Patterns: The session shifts to addressing how ontologies can be used in AI, such as enhancing LLM grounding, improving system reliability, and facilitating shared understanding across systems. The talk includes examples and lessons learned that can be applied to production.
• Building and Evaluating Ontologies: Discussion about constructing ontologies and quality metrics to assess their value. The process of implementing ontologies involves creating comprehensive and coherent descriptions of entities and relationships to form a usable knowledge graph.

Key Concepts

• Shared Understanding: Ontologies help build a shared understanding of concepts within different contexts, critical for interoperability across departments and systems.
• Declarative Representation: Ontologies formalize concepts like geographical regions, enabling more powerful query formulation and reasoning engines to work effectively.
• Entity Extraction and Resolution: Ontologies guide precise entity extraction and resolution, crucial for maintaining data quality and consistency.
• Inference and Reasoning: Using technologies like OWL, ontologies support logical inference and detection of inconsistencies within knowledge graphs.

Conclusion

Barrasa concludes by emphasizing the power of ontologies in enhancing the explicative, controllable, and reliable foundation for AI systems, supporting a knowledge-first approach to AI development and data management.

The session ends with the statement of having time for questions, encouraging further engagement on the topic.

This is the end of the AI-generated content.

Modern AI applications struggle not because of a lack of models, but because meaning is implicit, fragmented, and brittle. In this talk, we’ll explore how making semantics explicit (using ontologies and knowledge graphs) changes how we design, build, and operate AI systems. Drawing on real work from the GoingMeta.live podcast series, we’ll look at how ontologies move from theory to practice: grounding LLMs, improving reliability, enabling reasoning, and creating shared understanding across teams and systems. Expect concrete patterns, lessons learned, and examples you can apply to production AI today.