The Semantic Web is an approach to identifying, describing, linking and querying machine-processable data across application and organizational boundaries, based on the Resource Description Framework (RDF) data model. It extends document linking with structured descriptions that software can combine, rather than replacing human-readable Web pages.
The Semantic Web uses RDF for shared data representation, RDF Schema (RDFS) and the Web Ontology Language (OWL) for explicit vocabulary meanings, and SPARQL for querying. This means processing stated descriptions and logical relationships, not human-like understanding.
Identifiers and RDF Graphs
A URI (Uniform Resource Identifier) gives a resource, such as a person or document, a reusable name. Different publishers can use that name to add descriptions, but sharing an identifier does not ensure correct identity assignment or agreed meaning.
RDF statements are subject–predicate–object triples: the subject identifies what a statement concerns, the predicate identifies a property or relationship, and the object supplies a resource or value. In the graph formed by these triples, subjects and objects are nodes; each predicate labels a link directed from subject to object. Subjects and objects can also be blank nodes—unnamed graph nodes—and objects can be literals, such as text or numbers. Applications can exchange and combine RDF graphs even across differing schemas—ways of organizing data—but the common model does not resolve conflicting meanings or poor-quality data.
In a hypothetical graph, a URI-named book is linked to a URI-named person through a URI-named writtenBy relationship. A second triple gives the book a title, with “River” as its literal text value.
Vocabularies and Conditional Reasoning
RDFS describes classes, or types of resources, and properties and their relationships. If the example's writtenBy property is declared a subproperty of creator, an RDFS-aware application can infer that the same person is the book's creator. These declarations do not prescribe application behavior or automatically provide validation or readable presentation.
OWL expresses ontologies—formal descriptions of concepts and relationships—using computational logic. Supporting software can check expressed knowledge for consistency or derive implicit information, not establish real-world truth or automatically repair data. Both RDFS and OWL reasoning depend on the supplied facts and vocabulary declarations and on software applying the relevant semantics.
Looking for liquidity, exploring on-ramp/off-ramp services, or seeking expert guidance?
Get startedSPARQL Queries and Federation
SPARQL queries match patterns against available RDF graphs. A SELECT query fills query variables with matching RDF resources or values, often displayed in a table. An ordinary query need not return inferred answers unless the engine applies the relevant reasoning. SPARQL 1.1 also specifies RDF updates and protocols for query and update services.
For federated queries, an engine matches local patterns locally and delegates designated subqueries to other SPARQL endpoints—services that accept SPARQL queries. This does not automatically query arbitrary external databases or discover every relevant source.
Linked Data and Linked Open Data
Linked Data provides publishing and linking practices within the broader Semantic Web approach. Its principles call for naming things with HTTP URIs, providing useful RDF descriptions when those names are looked up, and including links to other URIs so people and applications can discover related descriptions. These are expectations for publishing behavior, not guarantees that descriptions or links will be correct.
Linked Data may be internal, personal or group-wide rather than public. Linked Open Data (LOD) additionally requires an open licence permitting free reuse; public accessibility alone is insufficient.
Core Semantic Web technologies and practices
| Technology or practice | Role | Reference | Reference status |
|---|---|---|---|
| RDF | RoleGraph data model | ReferenceW3C RDF specifications | Reference statusRecommendations and Working Group Notes |
| OWL | RoleLogic-based ontology language | ReferenceW3C OWL specifications | Reference statusNormative specification documents |
| SPARQL | RoleRDF queries and updates | ReferenceSPARQL 1.1 specifications | Reference statusW3C Recommendations |
| URI | RoleResource naming | ReferenceRFC 3986: URI syntax | Reference statusRFC cited by the relational-to-RDF Recommendation |
| Linked Data | RolePublishing and discovery | ReferenceTim Berners-Lee's Linked Data design note | Reference statusPersonal guidance |
Relational Data and Integration Limits
Relational database data can participate in the Semantic Web through a mapping from its data and schema to an RDF graph representation. A mapping can materialize a graph—create an actual RDF dataset—or define a virtual graph without copying all data into an RDF store. Either can be queried with SPARQL or traversed through an RDF graph API, a programming interface for accessing the graph. The relational database need not be replaced.
Vocabulary documentation complements formal declarations by explaining intended meanings. Publishers still need to coordinate identifier use and interpretation and address data quality; common formats and arbitrary tags alone do not establish shared semantics.
In a hypothetical version of the Linked Data design note's complete two-way browsing example, one RDF document describes a person and another a group. Both include the same statement that the person belongs to the group, so someone looking up either resource can discover that relationship. Keeping those copies consistent requires coordination across authors or programs. Automatic generation can support this pattern, while practical multi-source publication can make compromises: not every dataset must publish every incoming and outgoing relationship.
