Automated MEI Transcription of a Dataset of Electronic Drum Kit Recordings

abstract: We present a method for the automated transcription of human performances on electronic drum kits, captured in MIDI files, into music scores in the MEI encoding.

It works by parsing an input MIDI sequence into a tree-structured intermediate representation, using techniques from formal languages and NLP, post-processing the latter representation, with dedicated term rewriting rules, and exporting it into MEI. In an case study conducted on Magenta’s Groove MIDI Dataset, we successfully transcribed, without pre-training or manual adjustments, 333 MIDI files of this dataset, covering various genres, time signatures, tempi and of length up to 260 measures.

Recommended citation: Florent Jacquemard, Lydia Rodriguez-de la Nava. "Automated MEI Transcription of a Dataset of Electronic Drum Kit Recordings". In proceedings of the Music Encoding Conference (MEC), 2025.