Towards a (better) definition of the description of annotated MIR corpora

Today, annotated MIR corpora are provided by various research labs or companies, each one using its own annotation methodology, concept definitions, and formats. This is not an issue as such. However, the lack of descriptions of the methodology used - how the corpus was actually annotated, and by whom - and of the annotated concepts, i.e. what is actually described, is a problem with respect to the sustainability, usability, and sharing of the corpora. Experience shows that it is essential to define precisely how annotations are supplied and described. We propose here a survey and consolidation report on the nature of the annotated corpora used and shared in MIR, with proposals for the axis against which corpora can be described so to enable effective comparison and the inherent influence this has on tasks performed using them.