Data Sources & Credits
Most songs on this site come from published research datasets in which trained musicians transcribed the chords and song structure of the original recordings. We convert those transcriptions directly: the chord symbols are theirs, and the harmonic analysis on each page is generated from them by code, not written by AI.
A song is marked Verified when two independent expert transcriptions agree. Songs with a single expert transcription are labelled Expert transcription. Songs from lead sheets are labelled Community lead sheet: their chords are as written on the sheet, which describes the song as composed rather than any particular recording, and their section names are worked out by code from repeated phrases. The piano voicings are our playable arrangements of each chord symbol. We changed the original data by converting its chord notation, collapsing repeated chords and sections, and simplifying a small number of chords our player can't yet show.
Chord and structure transcriptions of 740 songs from the Billboard Hot 100 (1958–1991), made by trained musicians at McGill University.
J. A. Burgoyne, J. Wild & I. Fujinaga (2011). “An Expert Ground-Truth Set for Audio Chord Recognition and Music Analysis.” Proceedings of the 12th International Society for Music Information Retrieval Conference (ISMIR), 633–638.
Chord, key and structure annotations of the Beatles' studio albums and of recordings by Queen, Carole King, Michael Jackson and Zweieck, from the Centre for Digital Music at Queen Mary University of London.
C. Harte, M. Sandler, S. Abdallah & E. Gómez (2005). “Symbolic Representation of Musical Chords: A Proposed Syntax for Text Annotations.” ISMIR; M. Mauch et al. (2009). “OMRAS2 Metadata Project 2009.” ISMIR late-breaking session.
Two independent by-ear harmonic analyses (by Trevor de Clercq and David Temperley) of 200 songs from Rolling Stone's “500 Greatest Songs of All Time”. We publish a song only when the analysts substantially agree, and mark it verified when they agree on at least 90% of the song.
T. de Clercq & D. Temperley (2011). “A Corpus Analysis of Rock Harmony.” Popular Music 30(1), 47–70.
The collection through which we obtain all three datasets above, in a consistent format.
J. de Berardinis, A. Meroño-Peñuela, A. Poltronieri & V. Presutti (2023). “ChoCo: a Chord Corpus and a Data Transformation Workflow for Musical Harmony Knowledge Graphs.” Scientific Data 10, 641.
Lead sheets (melody with chord symbols) of public-domain songs, contributed by musicians to the former Wikifonia archive. We use only songs first performed in 1930 or earlier. Titles and songwriter credits are taken from each sheet itself, with spelling corrected; the collection's separate catalogue data (titles, names and dates from SecondHandSongs, CC BY-NC) is used only to check that a song is old enough and is not shown. Genres come from its Discogs data (CC0).
F. Simonetta, F. Carnovalini, N. Orio & A. Rodà (2018). “Symbolic Music Similarity through a Graph-based Representation.” Proceedings of Audio Mostly 2018 (AM’18), ACM. Zenodo, doi:10.5281/zenodo.4332855.
Artist genres used for browsing and filtering. We map each artist's Wikidata genres onto our own genre list, so a song's genres describe its artist rather than that particular recording.