Periscoped
Ontology Atlas
Definitions

What is an ontology, precisely?

There is no single sentence that captures it — the definition was built up over decades. Below is the canonical lineage, quoted from the primary sources, followed by a glossary of every core term. Everything here is cited; the full list is on the references page.

Defining “ontology”

The definition most of the field uses is the end of a chain of refinements. Read them in order — each fixes an ambiguity in the last.

The starting point is the notion of a conceptualization: “the objects, concepts, and other entities that are assumed to exist in some area of interest and the relationships that hold among them” — Genesereth & Nilsson, 1987. An ontology is a way of pinning that down.

An explicit specification of a conceptualization. Gruber, 1993 — the original and most-cited definition. “Explicit” means the concepts and constraints are enumerated deliberately, not left implicit.
A formal specification of a shared conceptualization. Borst, 1997 — adds formal (machine-readable, well-defined semantics) and shared (a consensus among a group, not one person's view).
A formal, explicit specification of a shared conceptualization. Studer, Benjamins & Fensel, 1998 — the synthesis of Gruber and Borst, and the definition most widely quoted today.
formalMachine-readable, with well-defined (logical) semantics.
explicitThe concept types and the constraints on their use are stated deliberately.
sharedConsensual knowledge, accepted by a group — not a private view.
conceptualizationThe abstract model of the relevant concepts and their relationships in a domain.
A logical theory accounting for the intended meaning of a formal vocabulary. Guarino, 1998 — reframes an ontology as a logical theory whose intended models approximate the conceptualization. This is the view that OWL later operationalized.
A formal explicit description of concepts in a domain of discourse (classes), properties of each concept (slots), and restrictions on slots (facets). Noy & McGuinness, 2001 — the practical, engineering-oriented definition from Ontology Development 101. They add: an ontology together with a set of individual instances of classes constitutes a knowledge base.
A Semantic Web language designed to represent rich and complex knowledge about things, groups of things, and relations between things … such that knowledge expressed in OWL can be reasoned with by computer programs. W3C OWL 2, 2009 — the operational standard. In OWL's terms an ontology is a formal description of a domain of interest, made of classes, properties, individuals, and data values.

The anatomy of an ontology

The building blocks, in the vocabulary of Noy & McGuinness and the W3C.

Class also: concept, type
A concept in the domain — a collection of objects that share properties (e.g. Person, Disease). Classes are the primary building blocks and are usually organized into a hierarchy.
Noy & McGuinness, 2001
Individual also: instance
A specific member of a class — the actual data (e.g. Ada is an individual of Person). An ontology plus its individuals is a knowledge base.
Noy & McGuinness, 2001
Property also: slot, role
A named relationship that describes or connects classes. Object properties link two individuals (owns); datatype properties attach a data value (hasAge); annotation properties carry metadata (labels, comments).
W3C OWL 2 · Noy & McGuinness, 2001
Facet also: restriction, role restriction
A constraint on a property: its cardinality (how many values), value type, and its domain (which class it applies to) and range (which values it may take).
Noy & McGuinness, 2001
Axiom
A logical statement asserted to be true in the ontology (e.g. “every Dog is a Mammal”). The axioms are what a reasoner operates on.
W3C OWL 2
Subsumption is-a, rdfs:subClassOf
The relationship where one class is a more specific kind of another (Dog ⊑ Mammal). Subclasses inherit the properties of their superclasses.
W3C RDF Schema
Taxonomy class hierarchy
The tree (or lattice) of classes connected by subsumption. The backbone of most ontologies, but not the whole of one — an ontology adds properties, facets, and axioms on top.
Noy & McGuinness, 2001

The languages

How ontologies are written down — the W3C Semantic Web stack.

RDF Resource Description Framework
The base data model: a graph of triples. An RDF triple is a 3-tuple (subject, predicate, object); a set of triples is an RDF graph. The W3C standard model for data interchange on the web.
W3C RDF 1.1
RDFS RDF Schema
The vocabulary layer over RDF: defines rdfs:Class, rdf:Property, and the core relations rdf:type, rdfs:subClassOf, rdfs:subPropertyOf, rdfs:domain, and rdfs:range.
W3C RDF Schema 1.1
OWL Web Ontology Language (OWL 2)
The expressive, logic-based ontology language for the Semantic Web. It adds the axioms RDFS cannot express — cardinality, disjointness, equivalence, property characteristics — with a formally defined meaning a reasoner can compute over. A W3C Recommendation since 2009.
W3C OWL 2
OWL 2 profiles EL, QL, RL
Restricted sublanguages of OWL 2 that trade expressivity for tractable (polynomial-time) reasoning — EL (large biomedical ontologies), QL (query answering over databases), RL (rule-based reasoning).
W3C OWL 2 Profiles
SKOS Simple Knowledge Organization System
A lighter-weight W3C model for thesauri, taxonomies, and controlled vocabularies as linked data. Where OWL models formal logic, SKOS captures “broader / narrower / related” term structures.
W3C SKOS Reference
SPARQL
The standard query language for RDF graphs — the “SQL of the Semantic Web.”
W3C SPARQL 1.1

Reasoning & logic

Why an ontology is a logical theory, not just a schema.

Description logic DL
A family of formal knowledge-representation languages — a decidable fragment of first-order logic — that provides the mathematical foundation for OWL. Concepts are unary predicates, roles are binary predicates.
Baader et al., 2003
SROIQ(D)
The specific description logic underlying OWL 2 DL. Highly expressive; its reasoning is decidable but computationally hard, which is why the tractable profiles exist.
Horrocks, Kutz & Sattler, 2006 · W3C OWL 2 Direct Semantics
TBox terminological box
The schema half of a knowledge base: the concepts, roles, and axioms defining how terms relate. Analogous to a database schema.
Description-logic convention
ABox assertional box
The data half: assertions about individuals (their types and relationships). Analogous to the rows in a database. TBox + ABox = the knowledge base.
Description-logic convention
Reasoner inference engine
Software (e.g. HermiT, ELK, Pellet) that computes entailments over an ontology: consistency (are the axioms contradiction-free?), subsumption (is A a subclass of B?), and instance checking (does x belong to class C?) — making implicit knowledge explicit.
W3C OWL 2
Open-world assumption OWA
The rule that what is not stated is unknown, not false. Ontologies reason under the OWA — a key difference from databases (which assume the closed world). Closed-world shape validation is done separately with SHACL.
Description-logic semantics

Practice & foundations

The terms you meet when building and reusing ontologies.

Competency question
A question the finished ontology must be able to answer. Written up front, competency questions define the scope and serve as the test of whether the ontology is done.
Noy & McGuinness, 2001
Upper (foundational) ontology
An ontology of the most general, domain-independent concepts — object, process, quality, relation. Domain ontologies anchor to one (BFO, DOLCE, SUMO) to stay mutually interoperable. BFO is standardized as ISO/IEC 21838-2.
Guarino, 1998 · ISO/IEC 21838
MIREOT
“Minimum Information to Reference an External Ontology Term” — the practice of importing just the specific terms you need from another ontology, plus their minimal context, rather than the whole thing.
Courtot et al., 2011
SHACL Shapes Constraint Language
A W3C standard for validating RDF graphs against a set of conditions (“shapes”) — closed-world data validation, complementary to OWL's open-world inference.
W3C SHACL
Knowledge graph
A graph-structured knowledge base that integrates data as entities and relationships, often governed by an ontology that gives the graph its schema and meaning.
common usage

Every definition here has a source.

See the full list of primary papers and W3C specifications — including Ontology Development 101 as a downloadable PDF.

Open the references