A corpus of frequently listened-to music in Germany over the past 50 years, enriched with metadata from online music databases
Year: 2026
Type: other
Abstract: This is POPTRAG, a corpus of 254,293 unique popular tracks released and frequently listened to in Germany over the past 50 years. The corpus combines four sources: 215,702 tracks from the weekly Media Control German single and album charts, comprising the 100 most relevant albums and songs per week in terms of physical and digital sales, streams, and broadcasts (18 November 1977–21 November 2024; reduced to 75% where linkage to Spotify identifiers was successful); 26,949 Spotify track recommendations from 126 German-market genre seeds (retrieved on 29 November 2024; median 325 per seed, IQR 25, range 49–343); 8,801 tracks from the 100 most prominently featured Spotify playlists for Germany (retrieved on 29 November 2024; median 90 tracks per playlist, IQR 42, range 28–750); and, finally, 11,994 unique track entries from Germany’s Daily Top 200 Spotify charts (from their launch in January 2017 until November 2024). We excluded tracks shorter than 30 seconds (853 tracks) and retained only unique title–artist combinations, preferring earlier release dates in cases of duplication. The remaining tracks were enriched with available metadata from Spotify, MusicBrainz, AcousticBrainz, Deezer, Discogs, and Genius. The resulting corpus comprises a total of 507 variables. POPTRAG is provided as tabular data in an .rds file. Respecting copyright, POPTRAG does not contain the full lyrics of the tracks. However, we calculated probabilities for lyrical expression types following the “More than Words” approach of Radke and Lepa (2026). These probabilities are provided as additional input features for machine-learning tasks.
Source: DepositOnce
Author: Radke, Markus
Language: English
Topic: Music and Audio Processing
Subfield: Signal Processing
Field: Computer Science
Domain: Physical Sciences
Open Access status: green
OpenAlex ID: W7204179586
