Dataset Card for SynTheory

Table of Contents

• Dataset Description
  • Dataset Summary
  • Supported Tasks and Leaderboards
  • Languages
  • How to use
• Dataset Structure
  • Data Instances
  • Data Fields
  • Data Splits
• Dataset Creation
  • Curation Rationale
  • Source Data
  • Annotations
  • Personal and Sensitive Information
• Considerations for Using the Data
  • Social Impact of Dataset
  • Discussion of Biases
  • Other Known Limitations
• Additional Information
  • Dataset Curators
  • Licensing Information
  • Citation Information
  • Contributions

Dataset Description

• Homepage: Do Music Generation Models Encode Music Theory?
• Repository: SynTheory
• Paper: Do Music Generation Models Encode Music Theory?

Dataset Summary

SynTheory is a synthetic dataset of music theory concepts, specifically rhythmic (tempos and time signatures) and tonal (notes, intervals, scales, chords, and chord progressions).

Each of these 7 concepts has its own config.

tempos consist of 161 total integer tempos (bpm) ranging from 50 BPM to 210 BPM (inclusive), 5 percussive instrument types (click_config_name), and 5 random start time offsets (offset_time).

time_signatures consist of 8 time signatures (time_signature), 5 percussive instrument types (click_config_name), 10 random start time offsets (offset_time), and 3 reverb levels (reverb_level). The 8 time signatures are 2/2, 2/4, 3/4, 3/8, 4/4, 6/8, 9/8, and 12/8.

notes consist of 12 pitch classes (root_note_name), 9 octaves (octave), and 92 instrument types (midi_program_name). The 12 pitch classes are C, C#, D, D#, E, F, F#, G, G#, A, A# and B.

intervals consist of 12 interval sizes (interval), 12 root notes (root_note_name), 92 instrument types (midi_program_name), and 3 play styles (play_style_name). The 12 intervals are minor 2nd, Major 2nd, minor 3rd, Major 3rd, Perfect 4th, Tritone, Perfect 5th, minor 6th, Major 6th, minor 7th, Major 7th, and Perfect octave.

scales consist of 7 modes (mode), 12 root notes (root_note_name), 92 instrument types (midi_program_name), and 2 play styles (play_style_name). The 7 modes are Ionian, Dorian, Phrygian, Lydian, Mixolydian, Aeolian, and Locrian.

chords consist of 4 chord quality (chord_type), 3 inversions (inversion), 12 root notes (root_note_name), and 92 instrument types (midi_program_name). The 4 chord quality types are major, minor, augmented, and diminished. The 3 inversions are root position, first inversion, and second inversion.

simple_progressions consist of 19 chord progressions (chord_progression), 12 root notes (key_note_name), and 92 instrument types (midi_program_name). The 19 chord progressions consist of 10 chord progressions in major mode and 9 in natural minor mode. The major mode chord progressions are (I–IV–V–I), (I–IV–vi–V), (I–V–vi–IV), (I–vi–IV–V), (ii–V–I–Vi), (IV–I–V–Vi), (IV–V–iii–Vi), (V–IV–I–V), (V–vi–IV–I), and (vi–IV–I–V). The natural minor mode chord progressions are (i–ii◦–v–i), (i–III–iv–i), (i–iv–v–i), (i–VI–III–VII), (i–VI–VII–i), (i–VI–VII–III), (i–VII–VI–IV), (iv–VII–i–i), and (VII–vi–VII–i).

Supported Tasks and Leaderboards

• audio-classification: This can be used towards music theory classification tasks.
• feature-extraction: Our samples can be fed into pretrained audio codecs to extract representations from the model, which can be further used for downstream MIR tasks.

How to use

The datasets library allows you to load and pre-process your dataset in pure Python, at scale. The dataset can be downloaded and prepared in one call to your local drive by using the load_dataset function.

For example, to download the notes config, simply specify the corresponding language config name (i.e., "notes"):

from datasets import load_dataset

notes = load_dataset("meganwei/syntheory", "notes")

Using the datasets library, you can also stream the dataset on-the-fly by adding a streaming=True argument to the load_dataset function call. Loading a dataset in streaming mode loads individual samples of the dataset at a time, rather than downloading the entire dataset to disk.

from datasets import load_dataset

notes = load_dataset("meganwei/syntheory", "notes", streaming=True)

print(next(iter(notes)))

Bonus: create a PyTorch dataloader directly with your own datasets (local/streamed).

Local:

from datasets import load_dataset
from torch.utils.data.sampler import BatchSampler, RandomSampler
from torch.utils.data import DataLoader

notes = load_dataset("meganwei/syntheory", "notes")
batch_sampler = BatchSampler(RandomSampler(notes), batch_size=32, drop_last=False)
dataloader = DataLoader(notes, batch_sampler=batch_sampler)

Streaming:

from datasets import load_dataset
from torch.utils.data import DataLoader

notes = load_dataset("meganwei/syntheory", "notes", streaming=True)
dataloader = DataLoader(notes, batch_size=32)

To find out more about loading and preparing audio datasets, head over to hf.co/blog/audio-datasets.

Example scripts

[More Information Needed]

Dataset Structure

Data Fields

[More Information Needed]

Dataset Creation

Curation Rationale

[More Information Needed]

Source Data

Initial Data Collection and Normalization

[More Information Needed]

Who are the source language producers?

[More Information Needed]

Annotations

Annotation process

[More Information Needed]

Who are the annotators?

[More Information Needed]

Personal and Sensitive Information

[More Information Needed]

Considerations for Using the Data

Social Impact of Dataset

[More Information Needed]

Discussion of Biases

[More Information Needed]

Other Known Limitations

For the notes music theory concept, there are 9,936 distinct note configurations. However, our dataset contains 9,848 non-silent samples. The 88 silent samples at extreme registers are unvoiceable with our soundfont. With a more complete soundfont, all 9,936 configurations are realizable to audio.

The silent samples are the following audio files: 0_0_C_10_Music_Box.wav, 0_0_C_56_Trumpet.wav, 0_0_C_68_Oboe.wav, 1_0_C#_10_Music_Box.wav, 1_0_C#_56_Trumpet.wav, 1_0_C#_68_Oboe.wav, 2_0_D_10_Music_Box.wav, 2_0_D_56_Trumpet.wav, 2_0_D_68_Oboe.wav, 3_0_D#_10_Music_Box.wav, 3_0_D#_56_Trumpet.wav, 3_0_D#_68_Oboe.wav, 4_0_E_10_Music_Box.wav, 4_0_E_56_Trumpet.wav, 4_0_E_68_Oboe.wav, 5_0_F_10_Music_Box.wav, 5_0_F_56_Trumpet.wav, 5_0_F_68_Oboe.wav, 6_0_F#_10_Music_Box.wav, 6_0_F#_56_Trumpet.wav, 6_0_F#_68_Oboe.wav, 7_0_G_10_Music_Box.wav, 7_0_G_56_Trumpet.wav, 7_0_G_68_Oboe.wav, 8_0_G#_10_Music_Box.wav, 8_0_G#_56_Trumpet.wav, 8_0_G#_68_Oboe.wav, 9_0_A_10_Music_Box.wav, 9_0_A_56_Trumpet.wav, 9_0_A_68_Oboe.wav, 10_0_A#_10_Music_Box.wav, 10_0_A#_56_Trumpet.wav, 10_0_A#_68_Oboe.wav, 11_0_B_10_Music_Box.wav, 11_0_B_56_Trumpet.wav, 11_0_B_68_Oboe.wav, 12_0_C_68_Oboe.wav, 13_0_C#_68_Oboe.wav, 14_0_D_68_Oboe.wav, 15_0_D#_68_Oboe.wav, 16_0_E_68_Oboe.wav, 17_0_F_68_Oboe.wav, 18_0_F#_68_Oboe.wav, 19_0_G_68_Oboe.wav, 20_0_G#_68_Oboe.wav, 21_0_A_68_Oboe.wav, 22_0_A#_68_Oboe.wav, 23_0_B_68_Oboe.wav, 24_0_C_68_Oboe.wav, 25_0_C#_68_Oboe.wav, 26_0_D_68_Oboe.wav, 27_0_D#_68_Oboe.wav, 28_0_E_68_Oboe.wav, 29_0_F_68_Oboe.wav, 30_0_F#_68_Oboe.wav, 31_0_G_68_Oboe.wav, 32_0_G#_68_Oboe.wav, 33_0_A_68_Oboe.wav, 34_0_A#_68_Oboe.wav, 35_0_B_68_Oboe.wav, 80_2_G#_67_Baritone_Sax.wav, 81_2_A_67_Baritone_Sax.wav, 82_2_A#_67_Baritone_Sax.wav, 83_2_B_67_Baritone_Sax.wav, 84_2_C_67_Baritone_Sax.wav, 85_2_C#_67_Baritone_Sax.wav, 86_2_D_67_Baritone_Sax.wav, 87_2_D#_67_Baritone_Sax.wav, 88_2_E_67_Baritone_Sax.wav, 89_2_F_67_Baritone_Sax.wav, 90_2_F#_67_Baritone_Sax.wav, 91_2_G_67_Baritone_Sax.wav, 92_2_G#_67_Baritone_Sax.wav, 93_2_A_67_Baritone_Sax.wav, 94_2_A#_67_Baritone_Sax.wav, 95_2_B_67_Baritone_Sax.wav, 96_2_C_67_Baritone_Sax.wav, 97_2_C#_67_Baritone_Sax.wav, 98_2_D_67_Baritone_Sax.wav, 99_2_D#_67_Baritone_Sax.wav, 100_2_E_67_Baritone_Sax.wav, 101_2_F_67_Baritone_Sax.wav, 102_2_F#_67_Baritone_Sax.wav, 103_2_G_67_Baritone_Sax.wav, 104_2_G#_67_Baritone_Sax.wav, 105_2_A_67_Baritone_Sax.wav, 106_2_A#_67_Baritone_Sax.wav, and 107_2_B_67_Baritone_Sax.wav.

Additional Information

Dataset Curators

[More Information Needed]

Licensing Information

[More Information Needed]

Citation Information

@inproceedings{Wei2024-music,
  title={Do Music Generation Models Encode Music Theory?},
  author={Wei, Megan and Freeman, Michael and Donahue, Chris and Sun, Chen},
  booktitle={International Society for Music Information Retrieval},
  year={2024}
}

Data Statistics