***Apologies for cross-posting***
We are glad to announce the release of the Multitrack Contrapuntal Music Archive (MCMA).
It is a small, yet high-quality and homogeneous dataset which was carefully compiled and edited specifically for the generation of counterpoint using neural machine translation [1-2]. MCMA is a symbolic dataset of pieces specifically collated and edited to comprise, for any given polyphonic work, independent parts. At this moment MCMA comprises 475 files, with the majority (239) being in three tracks, 153 in two tracks, and 83 in four or more tracks. All of the pieces currently in MCMA are from Baroque composers. The license type is CC BY 4.0.
We would like to open the dataset to contributions, in the hope that MCMA can continue to grow beyond our efforts. Detailed guidelines for contributors are available at:
https://mcma.readthedocs.io/en/latest/docs/guidelines.html
The repository is available here:
and the documentation here:
https://mcma.readthedocs.io/en/latest/
There is also another, forthcoming publication [3] in the pipeline regarding MCMA, which is expected to be published early 2021.
We hope that MCMA will be useful for some of your endeavours and we would be delighted to receive merge requests, and to integrate new collections to this corpus.
Sincerely,
Stefano, Eric, Gianluca, and Anna
____________________________________________________________________________
[1] Stefano Kalonaris, Thomas McLachlan and Anna Aljanaki (2020) Computational Linguistics Metrics for the Evaluation of Two-Part Counterpoint Generated with Neural Machine Translation. In Proceedings of the First Workshop on NLP for Music and Audio, Montreal, Canada. International Society for Music Information Retrieval Conference (ISMIR)