Ontologies: the semantic layer of knowledge graphs
An ontology is a formal, shared specification of the concepts in a domain and the relationships between them. In a knowledge graph it is the schema and rulebook: it says what kinds of things exist, how they relate, and what can be inferred. This page collects the standards, editors, reasoners, libraries, registries, reference ontologies and platforms used to build and apply ontologies.
What an ontology defines
Five building blocks recur in almost every ontology, whatever language or tool is used.
Classes
The kinds of things in the domain — Person, Drug, Invoice — arranged in a hierarchy of broader and narrower concepts.
Properties
The relationships between things, and the attributes they carry, with declared domains and ranges.
Individuals
The concrete instances that populate the classes; in a knowledge graph these are the nodes.
Axioms & Constraints
Logical statements and rules — disjointness, cardinality, equivalence — that give the model precise meaning.
Reasoning
Automated inference from the axioms: classifying concepts, detecting inconsistencies and materialising implied facts.
The ontology resource directory
Standards, editors, reasoners, libraries, registries, reference ontologies and platforms — search or browse below.
The W3C language for authoring ontologies: classes, properties, individuals and axioms with description-logic semantics, so a reasoner can infer new facts and check consistency.
Visit site →The lightweight vocabulary layer of RDF: classes, subclass and sub-property hierarchies, and property domains and ranges. Often all the schema a knowledge graph needs.
Visit site →A W3C model for publishing thesauri, taxonomies and controlled vocabularies as linked data, using concepts, broader/narrower links and labels.
Visit site →The W3C Shapes Constraint Language for validating RDF graphs against a set of shapes — the usual way to enforce data quality on top of an ontology.
Visit site →The W3C query language and protocol for RDF, used to query ontology-based data and, with entailment regimes, inferred triples.
Visit site →A JSON serialization of linked data whose context maps plain JSON keys onto ontology terms, bridging web APIs and knowledge graphs.
Visit site →A YAML-based modeling language for defining schemas and ontologies, with generators that produce OWL, SHACL, JSON Schema and other artifacts from one source.
Visit site →The free, open-source ontology editor and framework from Stanford. Supports OWL 2, reasoner integration and a large plugin ecosystem; the default tool for building ontologies.
Visit site →A web-based, collaborative ontology editor from Stanford, with shared editing, change tracking and discussion for ontology teams.
Visit site →A collaborative web application for editing OWL ontologies, SKOS thesauri and RDF datasets with workflow and validation support.
Visit site →A web application for visualizing ontologies as interactive node-link diagrams, useful for exploring and explaining ontology structure.
Visit site →An OWL 2 DL reasoner for Java that checks consistency, classifies ontologies and computes entailments.
Visit site →An open-source OWL 2 DL reasoner for Java, continuing the Pellet line, with Jena and OWL API integrations.
Visit site →A fast reasoner for the OWL 2 EL profile, designed for very large biomedical-style ontologies.
Visit site →The Java API and reference implementation for creating, manipulating and serializing OWL ontologies.
Visit site →A Java framework for semantic web and linked data applications: RDF and ontology APIs, SPARQL, rule and OWL inference, and the Fuseki server.
Visit site →A Python module for ontology-oriented programming: load OWL ontologies, manipulate classes and individuals as Python objects, and run reasoners.
Visit site →A Python library for working with RDF: parsing, serializing, graph manipulation and SPARQL queries.
Visit site →A command-line tool for automating ontology development workflows — merging, reasoning, templating, quality checks and release.
Visit site →A virtual knowledge graph system that maps relational databases to ontologies and rewrites SPARQL queries into SQL, without copying data.
Visit site →A large repository and browser of biomedical ontologies, with search, mappings and ontology metadata.
Visit site →A community of ontology developers committed to shared principles for interoperable, reusable life-science ontologies.
Visit site →A repository and browsing service for biomedical ontologies, with search and an API for looking up terms.
Visit site →A shared vocabulary of types and properties for structured data on the web, widely used in JSON-LD and search-engine markup.
Visit site →The Financial Industry Business Ontology from the EDM Council, defining shared concepts for financial instruments, entities and contracts.
Visit site →The widely used ontology of gene function, describing molecular functions, biological processes and cellular components.
Visit site →A top-level ontology used as a common upper layer by many domain ontologies, especially in biomedicine.
Visit site →A knowledge graph platform with OWL reasoning, SHACL constraint validation and virtual graphs over existing data sources.
Visit site →An RDF graph database with built-in OWL and RDFS reasoning, SPARQL support and semantic search.
Visit site →TopQuadrant's enterprise data governance platform for managing ontologies, taxonomies and shapes as shared, governed models.
Visit site →Semantic middleware for building and managing taxonomies and ontologies, and for using them to enrich content and knowledge graphs.
Visit site →How ontologies and knowledge graphs fit together
A knowledge graph stores facts as nodes and edges; an ontology gives those facts a shared vocabulary and logic. The two are usually built together.
Schema and data
The ontology is the schema layer — classes and properties — while the knowledge graph holds the instance data that conforms to it.
Inference
With an OWL or RDFS reasoner, a graph can answer questions that were never stated explicitly, such as inferring that an instance belongs to a broader class.
Validation
SHACL shapes check that incoming data matches the model, keeping a growing graph consistent.
Integration
Shared or reference ontologies let separate data sources be mapped onto one meaning, and tools like Ontop expose relational data through an ontology without copying it.
Property graphs too
Ontology thinking applies beyond RDF: property graph teams also define node and edge types, though usually with lighter, application-level schemas.
Context for AI
An ontology-backed knowledge graph gives language-model applications structured, explainable context — a common basis for GraphRAG.
Getting started
A practical path from a first ontology to a governed one.
| Step | Typical choice |
|---|---|
| Model the domain | Start from an existing reference ontology (Schema.org, FIBO, BFO, Gene Ontology) where one fits, and edit in Protégé or WebProtégé |
| Pick the language | RDFS for simple hierarchies, OWL 2 for richer axioms, SKOS for taxonomies and thesauri, LinkML if you want one source for several formats |
| Check and reason | HermiT or Openllet for OWL 2 DL, ELK for large EL ontologies |
| Validate the data | SHACL shapes against the instance graph |
| Automate builds | ROBOT for merging, reasoning and release; OWL API, Jena, Owlready2 or RDFLib in code |
| Store and query | An RDF store such as Stardog or GraphDB, queried with SPARQL |
| Publish and find | Register in BioPortal or the OBO Foundry for life-science work; publish with JSON-LD |