Capriccio-MIDI: Articulation-Aware Violin MIDI Tokens

Capriccio-MIDI is the symbolic component of the Capriccio dataset. It contains 49,243 violin-compatible musical segments represented as compound MIDI-token arrays with note-level articulation labels. Each array contains pitch, velocity, duration, inter-onset interval, within-bar position, bar index, and articulation tokens. This dataset accompanies the following paper: "VioLM: A Neural Language Model for Violin Synthesis with Articulation"Sunan Zou, Zhe Zhang, Yigitcan Özer, Guojie Luo, and Junichi YamagishiProceedings of the 27th International Society for Music Information Retrieval Conference (ISMIR), Abu Dhabi, UAE, 2026. Please cite this paper if you use Capriccio-MIDI: @inproceedings{ZouEtAl26_VioLM_ISMIR, author = {Sunan Zou and Zhe Zhang and Yigitcan {\"O}zer and Guojie Luo and Junichi Yamagishi}, title = {VioLM: A Neural Language Model for Violin Synthesis with Articulation}, booktitle = {Proceedings of the 27th International Society for Music Information Retrieval Conference ({ISMIR})}, address = {Abu Dhabi, UAE}, year = {2026} } The accompanying code is available at:https://github.com/nii-yamagishilab-visitors/VioLM The archive includes the tokenizer configuration, token-format and creation documentation, per-file SHA-256 checksums, validation tools, and a converter for reconstructing ordinary MIDI notes. The converter omits articulation-control events because standard MIDI does not define a universal articulation mapping. COPYINGCapriccio-MIDI is licensed under the Creative Commons Attribution-NonCommercial 4.0 International license (CC BY-NC 4.0). Please see LICENSE.txt for the license terms. The license applies only to rights that the licensors are authorized to grant. Users remain responsible for any rights in underlying compositions, arrangements, or MIDI transcriptions.