This article discusses the integration of semantic web technologies (ontology and inference) into audiovisual annotation based models and systems. The Advene project, aimed at all purpose hypervideo generation from annotated audiovisual documents, is used as a testbed. Advene principles and the Advene prototype are first presented, before a discussion on how ontology and reasoning have been easily integrated into the Advene framework. Some motivating examples are proposed, and our proposals and related works are discussed.