The W3C's Semantic Web initiative is based around the Resource Description Framework (RDF) and Web Ontology Language (OWL).
These declarative, model-oriented data languages have various serializations: RDF/XML is the interchange syntax. But more 'active' languages have been appearing, and the following come closest to traditional programming languages:
The "classic" language for noodling with SemWeb material is Notation 3, which is primarily used with the cwm data processor (originally written by Tim Berners-Lee):
"Fabl (pronounced "fable") is a native programming language for the Semantic Web. The Fabl object model is the RDF property graph, and the Fabl type system implements a subset of the OWL web ontology language...Fabl programs themselves are represented as Semantic Web objects. This representation of computation within the Semantic Web allows active content to be integrated seamlessly into RDF repositories, and provides a programming environment which simplifies the manipulation of RDF when compared to use of a conventional language via an API."
"Adenine is a Python-like language for manipulating RDF. In addition to being useful for general Semantic Web applications, Adenine also powers much of the user interface and agent framework in Haystack."
Haystack is a "universal information client".
"Adenosine is a language designed both to work on, and be distributed over, the semantic web. The language exploits the expressiveness of RDF whilst adopting a clean syntax based on a combination of Notation3 and ECMAScript."