Ontology Training: The Complete Course in Taxonomies, Knowledge Graphs and AI
🔒 This course requires registration
To access this course and all our learning materials, please register for the AI Fluency programme.
Register Now →Ontology training, end to end. This course teaches you to build, query, validate and govern ontologies and knowledge graphs, from the philosophical question of what a category is to the engineering question of why your reasoner is hanging.
It covers the W3C stack (RDF, RDFS, OWL 2, SPARQL, SHACL), the property graph world (Neo4j, Cypher, ISO GQL), the upper ontologies used in government and industry (BFO, DOLCE, gist, SUMO, BORO, HQDM, IES), the domain standards that carry real money (ISO 15926, IFC, Uniclass, FIBO, SNOMED CT, CIDOC-CRM), and the 2026 state of play on GraphRAG, neurosymbolic AI and agent memory.
45 lessons, a graded quiz on every lesson, and 25 workbook and checkpoint exercises that build one portfolio artifact. Free and open to access.
What you will learn
- Tell a taxonomy, a thesaurus, an ontology and a knowledge graph apart, and pick the right one for a given problem
- Write and read RDF, model with OWL 2, query with SPARQL, validate with SHACL and infer with SHACL-AF rules
- Choose an upper ontology on evidence rather than fashion, including BFO, UFO and gUFO, and 3D and 4D identity models
- Build crosswalks between standards with SSSOM and test them for falsifiability, not just consistency
- Judge when a knowledge graph genuinely beats vector retrieval, and when it does not
- Pick a development methodology, write competency questions that can be tested, and decide which layer carries the semantics
Course curriculum
45 lessons, each with a graded quiz.
Module 1: Foundations
What ontologies are, how they differ from taxonomies and thesauri, how identity and change are modelled, and the logic that makes inference possible.
- What Is an Ontology? From Aristotle to Enterprise AI
- Taxonomy vs Thesaurus vs Ontology vs Knowledge Graph
- Identity, Change and Time: 3D vs 4D Worldviews
- Logic for Ontologists: Open Worlds, Closed Worlds and Inference
Module 2: The semantic web stack
RDF, RDFS and OWL 2, SKOS for controlled vocabularies and the governance behind them, SPARQL, SHACL for validation and SHACL-AF for rules, reasoners, triplestores and SPARQL engines, and the bridge to property graphs and ISO GQL.
- RDF Explained: Triples, IRIs and Serializations
- RDFS and OWL 2: From Vocabularies to Description Logic
- SKOS: Modelling Controlled Vocabularies, Thesauri and Taxonomies
- Running a Controlled Vocabulary: Governance, Tooling and Quality
- SKOS or OWL: Choosing the Weakest Thing That Works
- SPARQL for People Who Know SQL
- SHACL: Validating Knowledge Graphs
- SHACL-AF Rules: Closing the Open World and Applying Defaults
- Node Expressions, SHACL Functions and Custom Targets
- Rules in Production: Provenance, Termination and Choosing Your Engine
- Reasoners in Practice: HermiT, ELK and Why Yours Hangs
- SPARQL Engines and Triplestores: Fuseki, Virtuoso, QLever and How to Choose
- Property Graphs and the RDF Bridge: Neo4j, Cypher and ISO GQL
Module 3: Method and governance
Competency questions and the methodologies behind them (101, NeOn, SAMOD, eXtreme Design, MOMo), where to load semantics, visual modelling tools, definitions and design patterns, OntoClean, the upper ontology shootout including UFO and gUFO, relators and power types, BORO, HQDM and IES, and testing ontologies in CI.
- Ontology Development 101: Competency Questions to First Release
- Ontology Engineering Methodologies: 101, NeOn, SAMOD, eXtreme Design and MOMo
- Competency Questions That Test Something: Types, Templates and CLaRO
- Seeing the Model: Visual Ontology Modelling and Visualisation Tools
- Definitions, Design Patterns and OntoClean
- Upper Ontologies Compared: BFO, DOLCE, gist, SUMO and 4D
- UFO, OntoUML and gUFO: The Foundational Ontology Conceptual Modelling Uses
- Beyond IsA and HasA: Relators, Power Types and Intensional Meaning
- BORO, HQDM and IES: 4D Ontology for Government and Defence
- Testing, CI and Governance: ROBOT, ODK and Falsifiability
Module 4: The standards atlas and crosswalks
How to read any data standard, and how to map between two of them with SSSOM and defensible mapping predicates.
- The Standards Atlas: How to Read Any Data Standard
- Crosswalks: SSSOM, Mapping Predicates and Ontology Alignment
Module 5: Domains, and who is buying
Defence, industry, construction, space, life sciences and food, finance, culture, and national programmes country by country.
- Defence and National Security: IES, HQDM and Allied Standards
- Industrial and Engineering: ISO 15926, IDO and the Industrial Ontology Foundry
- Construction: IFC, Uniclass, COBie and BOT
- Space: SSAO and Knowledge Graphs for Orbit
- Life Sciences and Food: OBO, GO, SNOMED CT, FoodOn and AGROVOC
- Finance and Commerce: FIBO, GS1 and schema.org
- Culture, Places and the Public Graph: CIDOC-CRM, GeoSPARQL and Wikidata
- National Programmes: Who Is Buying Ontology, Country by Country
Module 6: Ontologies in the LLM era
What LLMs actually changed, GraphRAG versus vector RAG, safe knowledge graph extraction, neurosymbolic systems, agent memory and MCP, and the 2026 research frontier.
- Did LLMs Kill the Semantic Web? What Actually Changed
- GraphRAG vs Vector RAG: When Knowledge Graphs Pay
- Building Knowledge Graphs from Text with LLMs, Safely
- Neurosymbolic AI: Neural Proposes, Symbolic Disposes
- Geometric Neurosymbolic AI: Box and Cone Embeddings of Ontologies
- Agent Memory and MCP: Knowledge Graphs for AI Agents
- The Research Frontier: Papers Shaping 2026
Module 7: Career and capstone
The tools, the portfolio and the route into knowledge graph engineering.
Frequently asked questions
What is an ontology in AI?
An ontology is a formal, machine-readable model of a domain: what kinds of thing exist, what properties they have, what relationships can hold between them, and what constraints must hold. It is normally written in OWL 2. The reason it matters for AI is that software can read it and reject a statement that contradicts it, which prose documentation cannot do.
What is the difference between a taxonomy and an ontology?
A taxonomy arranges terms in broader and narrower relations, and little else. An ontology adds the relationships and the formal axioms, so a reasoner can infer new facts and detect contradictions. A taxonomy can tell you that a spaniel is a dog. An ontology can tell you that a claim about that spaniel is impossible.
What is the difference between an ontology and a knowledge graph?
The ontology is the schema and the knowledge graph is the data. The ontology says what can be said in the domain: the classes, the properties, the constraints. The knowledge graph holds the actual entities and the actual claims about them. You can build a graph without an ontology, but then nothing can tell you the graph is wrong.
What is the difference between OWL and SHACL?
OWL 2 describes what is true in a domain and lets a reasoner infer more of it. SHACL checks whether a particular data graph conforms to a set of shapes. They answer different questions, so most real projects use both. The common mistake is treating SHACL as a sufficient correctness check when it only validates the shapes you wrote.
Can SHACL catch a term an AI model invented?
No, not on its own. SHACL operates under the open-world assumption, so it silently passes any property or class it has no shape for. In a published test across three vocabularies, open-world SHACL validated as conformant all 300 data graphs seeded with fabricated terms. A closed-world vocabulary gate caught all 300 with no false positives on clean data.
Do RAG systems and AI agents need an ontology?
They need something that can tell them they are wrong, and an ontology is the cheapest thing that does that job at the term level. Retrieval hands a model relevant text but gives it no way to detect that the entity it just asserted does not exist. The ontology plus a closed-world check is what rejects the bad triple before it enters the graph.
What is SPARQL used for?
SPARQL is the W3C query language for RDF data. Where SQL queries tables, SPARQL matches graph patterns across triples, so a single query can follow relationships several hops deep without writing joins. It supports SELECT for tabular results, CONSTRUCT for building new graphs, and ASK for yes or no checks. Anyone comfortable with SQL picks it up quickly.
What is GraphRAG?
GraphRAG retrieves over a knowledge graph rather than a flat vector index, so the model receives structured neighbours and relationship paths instead of loose passages. It performs better on multi-hop questions. It does not need an ontology to run, which is the trap: without a schema and a term-level check, it retrieves whatever the extraction step invented.
What is an upper ontology, and which ones are used in practice?
An upper ontology models domain-independent categories such as objects, events and time, so that specialised ontologies built on it can be aligned. BFO and DOLCE are the widely cited academic ones. In UK government and defence, the 4D upper ontologies IES and HQDM matter most, because they are the models real data standards are built against.
Is this ontology course free?
Yes. The course is open to access at no cost, including every lesson, every graded quiz and the workbook exercises. There is no payment step and no trial period.