boblsturm / aimusicgenerationchallenge2020

Repositories of systems submitted to the AI Music Generation Challenge 2020

The Ai Music Generation Challenge 2020

in the style of the Double Jigs in Captain Francis O’Neill’s “1001”

What?

Build an artificial system that generates the most plausible double jigs, as judged against the 365 published in F. O’Neill “The Dance Music of Ireland: O’Neill’s 1001” (1907). (See this cleaned and corrected version of the 1001 jigs: http://www.norbeck.nu/abc/book .) Up to two prizes will be awarded, and a performance of the best ones will occur at the 2020 Joint Conference on AI Music Creativity (CSMC+MuMe), Oct. 19-23 in Stockholm, Sweden (http://kth.se/aimusic2020). The panel of judges consists of four (human) experts in Irish traditional music and performance.

This challenge is part of a project that has received funding from the European Research Council (ERC) under the European Union’s Horizon 2020 research and innovation programme (Grant agreement No. 864189 MUSAiC: Music at the Frontiers of Artificial Creativity and Criticism).

Why?

This challenge has three aims:

1. to promote meaningful approaches to evaluating music Ai;
2. to see how music Ai research can benefit from traditional music, and how traditional music can benefit from music Ai research;
3. to facilitate discussions about the ethics of music Ai research applied to traditional music practices.

How?

1. BEFORE JULY 17, register your intent to participate by notifying the organizer.
2. Build a music AI that generates music. You can train your AI on anything, but remember that the results will be judged against the 365 double jigs in O’Neill’s “1001”. To facilitate judging, the music you submit must be rendered either as ABC notation, staff notation, MIDI, or mp3-compressed audio files.
3. Have your AI generate 10,000 tunes.
4. Write a brief technical document describing how you built your system, presenting some of its features and outcomes, and linking to your code and models for reproducibility.
5. BEFORE SEPTEMBER 22, email the organizer a link to download your generated collection and your technical document.
6. Only one submission from each participant will be allowed, so choose your best one.

Evaluation

For each submitted collection, five tunes will be selected at random. For example, given an ordered collection of 10,000 music files, five random integers will be drawn without replacement in [1,10000]. The files indexed by those numbers will be selected. The tunes will be compiled into a collection (random ordering), which will then be assessed by each judge. Information relating to the participating AI will not be disclosed.

Stage 1 (judges working individually):

• For each tune, the judge will gauge whether it is plagiarized. If it is, then the tune will not be considered further.
• Otherwise, the judge will gauge whether the rhythm is close to that of a double jig. If it is not, then the tune will not be considered further.
• Otherwise, the judge will determine whether the pitch range of the tune is characteristic. If it is not, then the tune will not be considered further.
• Otherwise, the judge will determine whether the mode and accidentals are characteristic. If it is not, then the tune will not be considered further.

Stage 2 (judges working individually):

• For each tune that has passed through Stage 1, the judge will rate it along multiple dimensions with reference to the double jigs in O’Neill’s “1001”: i. melody ii. structure and coherence iii. playability on traditional Irish instruments iv. memorability v. interestingness.
• Each judge will send to the organizer a report of their assessments.

Stage 3 (judges working together):

• Each judge will present to the other judges the best tunes from their collections.
• Together they will decide which is the best double jig (or to award no prize).

Stage 4 (judges working together with organizer):

• The participating music AI systems that have produced highly ranked double jigs will be subject to an analysis of their consistency of quality.
• The organizer and judges will query the submitted collections of these AI by selecting tunes at random and assessing their quality.
• The judges together will decide which AI is most consistent in producing double jigs of high quality (or to award no prize).

Results

• Results are described here.
• A brief documentary of the 2020 Challenge is here.
• See this article: B. L. T. Sturm and H. Maruri-Aguilar, “The Ai Music Generation Challenge 2020: Double jigs in the style of O’Neill’s “1001”,” Journal of Creative Music Systems, 2021.