Web 2.0 technologies can be considered a loosely defined set of Web application styles that foster a kind of media consumer more engaged, and usually active in creating and maintaining Internet contents. Thus, Web 2.0 applications have resulted in increased user participation and massive user-generated (or user-published) open multimedia content, some of which is potentially useful for education. In this context, the problem from the educator's perspective is selecting and annotating existing content so that software applications can filter fragments that were previously marked as useful for particular learning needs. This article discusses a solution for this problem that is non-intrusive to existing applications. This solution fits in the philosophy of multiple metadata profiles, allows for expressing fine-grained learning needs, and leverages the growing mass of contents by reusing well-established domain ontologies. A description of the technical aspects required for the infrastructure supporting such solution is first provided. Then, the solution is contextualized with a case study using a knowledge model of gene-related elements--the Gene Ontology--to semantically annotate videos in "YouTube" that could be used in teaching biology and bioinformatics. (Contains 12 notes and 2 figures.)