---
language:
- en
license: cc-by-4.0
dataset_info:
  features:
  - name: id
    dtype: string
  - name: start_time
    dtype: int32
  - name: question
    dtype: string
  - name: answer
    dtype: string
  splits:
  - name: train
    num_bytes: 13432031
    num_examples: 5802
  download_size: 3860760
  dataset_size: 13432031
configs:
- config_name: default
  data_files:
  - split: train
    path: data/train-*
---

# Maestro ABC Notation 25s Dataset

## Dataset Summary

This is based on V3.0.0 of the Maestro dataset.

The **Maestro ABC Notation 25s Dataset** is a curated collection of question-and-answer pairs derived from short audio clips within the [MAESTRO dataset](https://magenta.tensorflow.org/datasets/maestro). Each entry in the dataset includes:

- An `id` corresponding to the original audio file.
- A `start_time` marking where the 25-second audio clip begins within the full track.
- A `question` designed to prompt music transcription in ABC notation.
- An `answer` that provides the transcription in ABC notation format.

This dataset is crafted for training multi-modal audio-language models (such as [Spotify Llark](https://research.atspotify.com/2023/10/llark-a-multimodal-foundation-model-for-music/) and [Qwen2-Audio](https://github.com/QwenLM/Qwen2-Audio)) with a focus on music transcription tasks. The MIDI-to-ABC conversion is achieved with a modified script based on [this code](https://github.com/jwdj/EasyABC/blob/master/midi2abc.py).

### Why ABC?

The reasons for choosing this notation are:

- It's a minimalist format for writing music
- It's widely used and popular, language models already have good comprehension and know a lot about ABC notation.
- It's flexible and can easily be extended to include tempo changes, time signature changes, additional playing styles like mentioned above, etc…

### Dataset Modifications to ABC Format

- Default octaves have been assigned to each instrument, using their most commonly played range. This reduces redundant octave notation.
- For consistency, I excluded pieces that contain time signature changes or significant tempo variations (greater than 10 BPM).
- All samples in this dataset contain active musical parts - sections with complete silence have been removed.

## Licensing Information

- **MAESTRO Dataset**: The audio files are sourced from the MAESTRO dataset, licensed under the Creative Commons Attribution Non-Commercial Share-Alike 4.0 license. Please refer to the [MAESTRO dataset page](https://magenta.tensorflow.org/datasets/maestro) for full licensing details.

## Citation Information

If you utilize this dataset, please cite it as follows:

```bibtex
@dataset{maestro_abc_notation_25s_2024,
  title={MAESTRO ABC Notation Dataset},
  author={Jon Flynn},
  year={2024},
  howpublished={\url{https://huggingface.co./datasets/jonflynn/maestro_abc_notation_25s}},
  note={ABC notation for the MAESTRO dataset split into 25-second segments},
}
```

For the original MAESTRO dataset, please cite the following:

```bibtex
@inproceedings{hawthorne2018enabling,
  title={Enabling Factorized Piano Music Modeling and Generation with the {MAESTRO} Dataset},
  author={Curtis Hawthorne and Andriy Stasyuk and Adam Roberts and Ian Simon and Cheng-Zhi Anna Huang and Sander Dieleman and Erich Elsen and Jesse Engel and Douglas Eck},
  booktitle={International Conference on Learning Representations},
  year={2019},
  url={https://openreview.net/forum?id=r1lYRjC9F7},
}
```