Datasets:

NaturNestAI
/

electronic-music-knowledge

license: cc0-1.0
task_categories:
  - text-classification
  - feature-extraction
language:
  - en
tags:
  - music
  - electronic-music
  - dj
  - music-information-retrieval
  - knowledge-graph
  - genre-classification
  - discogs
  - ishkur
pretty_name: Electronic Music Knowledge
size_categories:
  - 10M<n<100M
configs:
  - config_name: tracks
    data_files: tracks/train-*.parquet
  - config_name: artists
    data_files: artists/train-*.parquet
  - config_name: labels
    data_files: labels/train-*.parquet
  - config_name: genres
    data_files: genres/train-*.parquet
  - config_name: genre_graph
    data_files: genre_graph/train-*.parquet
default_config_name: tracks

Electronic Music Knowledge

The largest open electronic music metadata dataset. 18.3M tracks, 1.4M artists, 353K labels, 832 genres with evolution graph.

Built for DJ Treta — an autonomous AI DJ — but useful for any music AI research.

Dataset Summary

Config	Rows	Description
`tracks`	18,315,675	Electronic music tracks with title, artist, genre/style, label, year, country
`artists`	1,424,582	Artists with primary genres, labels, country, active years, track count
`labels`	352,984	Record labels with genres, country, founding year
`genres`	832	Electronic genre taxonomy from Ishkur's Guide + Discogs (166 with BPM ranges)
`genre_graph`	352	Genre evolution relationships with time ranges

Quick Start

from datasets import load_dataset

# Load tracks (default config)
tracks = load_dataset("NaturNestAI/electronic-music-knowledge", "tracks", split="train")

# Load other configs
artists = load_dataset("NaturNestAI/electronic-music-knowledge", "artists", split="train")
genres = load_dataset("NaturNestAI/electronic-music-knowledge", "genres", split="train")
labels = load_dataset("NaturNestAI/electronic-music-knowledge", "labels", split="train")
graph = load_dataset("NaturNestAI/electronic-music-knowledge", "genre_graph", split="train")

Examples

Find melodic techno tracks

tracks = load_dataset("NaturNestAI/electronic-music-knowledge", "tracks", split="train")
melodic = tracks.filter(lambda x: x["subgenre"] == "Melodic House & Techno")
print(f"{len(melodic)} melodic techno tracks")

Find artists on a label

artists = load_dataset("NaturNestAI/electronic-music-knowledge", "artists", split="train")
drumcode = artists.filter(lambda x: x["labels"] and "Drumcode" in str(x["labels"]))

Genre evolution graph

graph = load_dataset("NaturNestAI/electronic-music-knowledge", "genre_graph", split="train")
# What influenced a genre?
influences = graph.filter(lambda x: x["target_genre"] == "melodictechno")

Schema

tracks

Column	Type	Description
`id`	string	Unique track identifier
`title`	string	Track/release title
`artist_name`	string	Artist name
`discogs_artist_id`	string	Discogs artist ID
`discogs_release_id`	string	Discogs release ID
`subgenre`	string	Primary Discogs style (e.g., "Melodic House & Techno")
`styles_json`	string	JSON array of all Discogs styles
`label`	string	Record label name
`country`	string	Release country
`year`	int	Release year
`search_query`	string	Pre-computed "Artist - Title" for YouTube/music search
`source`	string	Data source ("discogs" or "ishkur")

artists

Column	Type	Description
`id`	string	"discogs:{artist_id}"
`name`	string	Artist name
`primary_genres`	string	Primary genre/style
`labels`	string	Known label
`country`	string	Country of origin
`active_since`	int	Year of earliest release
`track_count`	int	Number of tracks in dataset

genres (Ishkur subset has BPM ranges)

Column	Type	Description
`id`	string	Genre slug
`name`	string	Genre name
`scene`	string	Ishkur scene grouping (House, Techno, Trance, etc.)
`bpm_low` / `bpm_high`	int	Typical BPM range
`energy_typical`	int	Typical energy level (1-10)
`aliases`	string	Alternative genre names

labels

Column	Type	Description
`id`	string	"discogs:{label_id}"
`name`	string	Label name
`primary_genres`	string	Primary genre
`country`	string	Country
`founded_year`	int	Year of earliest release

genre_graph

Column	Type	Description
`source_genre`	string	Genre that influenced
`target_genre`	string	Genre that was influenced
`start_year` / `end_year`	int	Time range of influence

Data Sources

Source	License	Contribution
Discogs Data Dump (April 2026)	CC0 1.0	4.9M electronic releases, 1.4M artists, 353K labels, 666 styles
Ishkur's Guide to Electronic Music v3	Open	166 genre taxonomy with BPM ranges, 11K tracks, evolution graph

Planned Enrichment (v2)

BPM and musical key from AcousticBrainz (29.5M tracks, CC0)
Artist similarity graph
MusicBrainz cross-reference IDs
DJ set transition data

Pipeline

Built with VeltriaAI/music-intelligence — extensible source adapter architecture. Add new data sources by dropping a Python file.

Citation

@dataset{electronic_music_knowledge_2026,
  title={Electronic Music Knowledge},
  author={NaturNest AI},
  year={2026},
  url={https://huggingface.co/datasets/NaturNestAI/electronic-music-knowledge},
  license={CC0-1.0}
}