RDF is the culmination of an initiative, proposed in 1994 by Tim Berners Lee, to formalize metadata. It relies on an XML syntax and a predictive syntax (called « N3 ») that can both be considered as a graph serialization format. An RDF document is a set of triplets (subject, predicate, object) that represent a resource (the subject) by assigning a value (the object) to one of its properties (the predicate). RDF triples are based on the subject-verb-object structure of the simplest expressions of natural language. A relation called subsumption can be defined between two RDF documents. When it is established as a rdf document a subsumes b, it signifies that the situation represented by b is a particular case of that represented by a.

It may be necessary to organize the vocabulary used in an RDF document in the form of a rudimentary ontology. To do this, we have the RDFS language, which makes it possible to define classes and properties; and structure them using a standardized terminology of primitives (classes Class, Property, etc., and properties type, subClassOf, subPropertyOf, etc.). This terminology has the advantage of being intuitive and synthetic but can sometimes be insufficient. In 2003, a new language much more powerful than RDFS appears: the OWL language. In particular, it makes it possible to overcome certain shortcomings of RDFS.

The RDF / S language (RDF / S stands for RDF or RDFS) has a formal semantics defined in model theory that states the mathematical meaning of the objects in an RDF / S document and the subsumption relation that may exist between them. objects. This language is therefore among the languages ​​of knowledge representation offering a robust theoretical support that can be the subject of mathematically verifiable theorems.