In this paper we present the music information plane and the dfferent levels of information extraction that exist in the musical domain. Based on this approach we propose a way to overcome the existing semantic gap in the music field. Our approximation is twofold: we propose a set of music descriptors that can automatically be extracted from the audio signals, and a top-down approach that adds explicit and formal semantics to these annotations. These music descriptors are generated in two ways: as derivations and combinations of lower-level descriptors and as generalizations induced from manually annotated databases by the intensive application of machine learning. We belive that merging both approaches (bottom-up and top-down) can overcome the existing semantic gap in the musical domain.

The reported research has been funded by the EU-FP6-IST-507142 project SIMAC (Semantic Interaction with Music Audio Contents).