MUCOSA (Music Content Semantic Annotator) is an environment for the annotation and generation of music at different levels of abstraction. It is composed of three tiers: an annotation client that deals with microannotations (i.e. within-file annotations), a collection tagger, which deals with macro-annotations (i.e. acrossfiles annotations), and a collaborative annotation subsystem, which manages large-scale annotation tasks that can be shared among different research centres. The annotation client is an enhanced version of WaveSurfer, a speech annotation tool. The collection tagger includes tools for automatic generation of unary descriptors, invention of new descriptors, and propagation of descriptors across sub-collections or playlists. Finally, the collaborative annotation subsystem, based on Plone, makes possible to share the annotation chores and results between several research institutions. A collection of annotated songs is available, as a “starter pack” to all the indivi...