ABSTRACT

One of the key challenges in engineering is to gather large sets of relevant, representative data. This chapter discusses a number of key datasets for Intelligent Music Production and shows how they address some of the main challenges in the field. It presents a number of methods for collecting this information, and demonstrate the complexity of gathering useful, representative data. A range of useful data stores have been tried and tested in the field of music production. The Mix Evaluation Dataset contains around 20 multitrack recordings, most of them freely available. Source separation datasets are inherently useful for music production due to their reliance on measuring original stems against signals taken from decomposed mixes. Other than multitrack audio and annotated production data, an extensive range of relevant datasets are available from a wide number fields. One of the more common types of sound production data is recorded music. The Music Ontology provides an architecture for the representation of musical data.