Ontologies and RDFS/OWL

What Is an Ontology?

An ontology is a formal specification of a domain's vocabulary — the classes (types of things), properties (relationships), and the rules governing them. If the knowledge graph is the data, the ontology is the schema that gives that data meaning.

Think of an ontology as a shared dictionary for your domain. Without it, "Person" in one dataset might mean something different from "Person" in another. With a shared ontology, machines can understand and interoperate across datasets.

The Two Levels of a Knowledge Graph

Layer	Name	Content
T-Box	Terminological Box	Classes, properties, axioms (the schema)
A-Box	Assertional Box	Individual instances and their relationships (the data)

The T-Box defines what kinds of things exist. The A-Box says what things actually exist.

RDFS: RDF Schema

RDFS (RDF Schema) is the simplest ontology language. It provides:

• rdfs:Class — declares a class
• rdfs:subClassOf — declares class inheritance
• rdf:Property — declares a property
• rdfs:domain — specifies which class a property applies to
• rdfs:range — specifies which class/datatype a property's values must be
• rdfs:label / rdfs:comment — human-readable names and descriptions

RDFS is intentionally simple. For more expressive power, you need OWL.

@prefix rdfs: <http://www.w3.org/2000/01/rdf-schema#> .
@prefix rdf:  <http://www.w3.org/1999/02/22-rdf-syntax-ns#> .
@prefix xsd:  <http://www.w3.org/2001/XMLSchema#> .
@prefix ex:   <http://example.org/schema/> .

# Class definitions
ex:Agent     a rdfs:Class ;
    rdfs:label "Agent" .

ex:Person    a rdfs:Class ;
    rdfs:subClassOf ex:Agent ;
    rdfs:label "Person" ;
    rdfs:comment "A human individual." .

ex:Employee  a rdfs:Class ;
    rdfs:subClassOf ex:Person ;
    rdfs:label "Employee" .

ex:Organization  a rdfs:Class ;
    rdfs:subClassOf ex:Agent ;
    rdfs:label "Organization" .

# Property definitions
ex:name  a rdf:Property ;
    rdfs:domain ex:Agent ;
    rdfs:range  xsd:string ;
    rdfs:label  "name" .

ex:age  a rdf:Property ;
    rdfs:domain ex:Person ;
    rdfs:range  xsd:integer ;
    rdfs:label  "age" .

ex:worksFor  a rdf:Property ;
    rdfs:domain ex:Employee ;
    rdfs:range  ex:Organization ;
    rdfs:label  "works for" .

OWL: Web Ontology Language

OWL (Web Ontology Language) adds much richer expressiveness on top of RDFS:

OWL Construct	Meaning
owl:equivalentClass	Two classes have the same members
owl:disjointWith	No instance can be in both classes
owl:inverseOf	A property is the inverse of another
owl:TransitiveProperty	If A→B→C then A→C
owl:FunctionalProperty	Each subject has at most one value
owl:cardinality	Exact number of values required

OWL is the basis for reasoning — an OWL reasoner can automatically infer new facts from existing ones (covered in Chapter 7).

Well-Known Ontologies

You don't have to write ontologies from scratch. Many reusable vocabularies exist:

• FOAF (Friend of a Friend) — for people and social networks
• Schema.org — Google's vocabulary for web content; covers products, events, reviews, etc.
• Dublin Core — for metadata like title, creator, date
• PROV-O — for provenance and data lineage
• Gene Ontology — for biological processes, functions, and components

@prefix owl:  <http://www.w3.org/2002/07/owl#> .
@prefix rdfs: <http://www.w3.org/2000/01/rdf-schema#> .
@prefix ex:   <http://example.org/schema/> .

# Inverse properties
ex:worksFor   owl:inverseOf  ex:employs .

# Transitive property
ex:locatedIn  a owl:TransitiveProperty .
# → if London locatedIn England, England locatedIn UK
#   then London locatedIn UK (inferred)

# Functional property — a person has exactly one birth date
ex:birthDate  a owl:FunctionalProperty .

# Disjoint classes — nothing can be both a Person and an Organization
ex:Person  owl:disjointWith  ex:Organization .

# Equivalent classes — ex:Employee ≡ schema:Employee
ex:Employee  owl:equivalentClass  <https://schema.org/Employee> .

# Cardinality restriction — a Company must have at least 1 employee
ex:Company  rdfs:subClassOf [
    a owl:Restriction ;
    owl:onProperty ex:employs ;
    owl:minCardinality 1
] .

Summary

In this chapter you learned:

1. An ontology defines the T-Box (schema) for your knowledge graph — classes, properties, and axioms
2. RDFS provides simple class hierarchies and property constraints
3. OWL adds richer semantics: inverse properties, transitivity, cardinality, disjoint classes
4. The Open World Assumption means missing facts are unknown, not false

Next, we'll learn how to query a knowledge graph using SPARQL.