This paper proposes to apply the RDF framework to the representation of linguistic annotations. We argue that RDF is a suitable data model to capture multiple annotations on the same text segment, and to integrate multiple layers of annotations. Besides the idea of using RDF for this purpose, the main contribution of the paper is an OWL ontology, called TELIX (Text Encoding and Linguistic Information eXchange), which models annotation content. This ontology builds on the SKOSXL vocabulary, a W3C standard for lexical entities representation as RDF graphs. We extend SKOS in order to capture lexical relations between words (e.g., synonymy), as well as to support word sense disambiguation, morphological features and syntactic analysis, among others. Additionally, a formal mapping of feature structures to RDF graphs is defined, enabling complex composition of linguistic entities. Finally, the paper also suggests the use of RDFa as a convenient syntax that combines source texts and linguistic annotations in the same file.