Skip to content
Typedown

RDF: Meaning as Triples

On this page

RDF (Resource Description Framework) is a framework for expressing machine-readable information about resources (can be anything as defined like the URI spec). RDF enables automated processing and exchange of information across applications without losing meaning.

In the Semantic Web stack, XML provides structure but its tags have no machine-interpretable meaning: <creator> is just a string to an XML parser. RDF adds a universal data model (triples with IRIs) on top, so every relationship and resource has a globally unique identifier. This makes data interoperable and mergeable across sources. In that sense, RDF is metadata about resources, described in a way machines can reason about.

Remark: Think of it like this - With XML, data can be displayed to human & human can write programs to process the XML documents. Interpretation of XML is still under human's discretion. With RDF, there's a standard way to interpret the XML documents.

RDF is another graph model like property graphs. For how RDF compares to property graphs, see The Graph Space.

RDF is recommended by W3C.

  • Enriching web pages with machine-readable metadata (schema.org).

  • Linking datasets across sources (e.g. paintings to artist information in Wikidata).

  • Standards-compliant database data exchange.

  • Cross-dataset querying via SPARQL.

Subpages:

  • Data Model: Triples, RDF terms, named graphs, semantics.

  • Serialization Formats: Turtle, N-Triples, TriG, N-Quads, JSON-LD, RDFa, RDF/XML.

  • RDFS: Vocabularies and class hierarchies for RDF.

See also OWL for full ontology language and SPARQL for querying.

Semantics of RDF Graphs ​

Motivation:

  • Syntax alone is not enough to reliably merge data across sources.

  • Two datasets can use the same IRI to mean different things, or different IRIs to mean the same thing.

RDF addresses this with a model-theoretic semantics (RDF Semantics):

  • Every IRI denotes some thing in the world (its referent).

  • A triple S P O is true when the referent of S and the referent of O are actually related by the relation denoted by P.

  • A graph is true when all its triples are true.

That is all plain RDF semantics defines, without inference rules & class hierarchies, just truth conditions.

This foundation lets software check consistency and draw conclusions. A reasoner can derive triples that are logically implied by existing ones (entailment), or detect contradictions. For example, ex:bob ex:age "forty"^^xsd:integer . is inconsistent because "forty" does not satisfy xsd:integer.

RDFS and OWL build inference rules on top of this, but are not part of RDF's core semantics. Each introduces privileged IRIs whose meaning is hardcoded into processors:

  • RDFS: adds rules via rdfs:domain, rdfs:subClassOf, etc. For example, foaf:knows rdfs:domain foaf:Person . lets a reasoner derive ex:bob rdf:type foaf:Person . from ex:bob foaf:knows ex:alice .

  • OWL: adds richer rules (cardinality, disjointness, equivalence) on top of RDFS.

Axioms are not written in RDF. They are formal rules in the spec, implemented in processor code. What you write in RDF are vocabulary triples that act as inputs to those rules:

  • The axiom "if P rdfs:domain C and X P Y, infer X rdf:type C" lives in the spec.

  • The triple foaf:knows rdfs:domain foaf:Person . is a normal RDF triple that triggers it.

RDF Schema also draws no hard line between classes and instances: the same IRI can serve both roles at once.

Linked Data ​

RDF data is widely published on the Web as Linked Data: datasets connected by shared IRIs and queryable via SPARQL.

Notable examples include Wikidata, DBpedia, WordNet, Europeana, and VIAF.

Different publishers often mint different IRIs for the same entity. owl:sameAs lets anyone assert that two IRIs refer to the same resource, so processors can merge data across datasets:

turtle
<http://dbpedia.org/resource/Leonardo_da_Vinci>
    owl:sameAs <http://viaf.org/viaf/24604287/> .

Resources ​

Specs ​

  • W3C RDF 1.2 Primer is a gentle intro with examples (Warning: It's an informative Group Note Drafts and no guarantee is made regarding its claims).

  • W3C RDF 1.2 Concepts is the formal abstract data model. This is a normative specs, as oppose to the above specs.

Others ​