CITATION — REFERENCE ENTRY
Chord Recognition in Symbolic Music: A Segmental CRF Model, Segment-Level Features, and Comparative Evaluations on Classical and Popular Music — Transactions of the International Society for Music Information Retrieval
- Key
- masada2019-symbolic-chord-recognition
- Authors
- Masada, Kristen; Bunescu, Razvan
- Issued
- 2019
- Type
- article-journal
- Container
- Transactions of the International Society for Music Information Retrieval
- Volume
- 2
- Issue
- 1
- Pages
- 1-13
Raw CSL JSON
{
"DOI": "10.5334/tismir.18",
"URL": "https://doi.org/10.5334/tismir.18",
"page": "1-13",
"type": "article-journal",
"issue": "1",
"title": "Chord Recognition in Symbolic Music: A Segmental CRF Model, Segment-Level Features, and Comparative Evaluations on Classical and Popular Music",
"author": [
{
"given": "Kristen",
"family": "Masada"
},
{
"given": "Razvan",
"family": "Bunescu"
}
],
"issued": {
"date-parts": [
[
2019
]
]
},
"volume": "2",
"container-title": "Transactions of the International Society for Music Information Retrieval"
}
Claims
-
Masada and Bunescu evaluated on four corpora: 60 Bach chorales, 27 Mozart and Beethoven theme-and-variation sets, excerpts from Kostka and Payne's theory textbook, and 59 pop and rock songs.
"this is the Bach Choral Harmony Dataset, a corpus of 60 four-part Bach chorales"
-
Chord changes tend to fall on strong metrical positions, passing and neighbouring tones on weak ones, and suspensions on strong beats.
"chord changes tend to happen in strong metrical positions; figurations such as passing and neighboring tones appear in metrically weak positions, whereas suspensions appear on metrically strong beats."
-
Approaches that tag short spans of music cannot use segment-level features, which are needed, for example, to identify figuration notes or to label segments that do not start with the chord's root.
"A significant drawback of these short-span tagging approaches is that they do not explicitly model candidate segments during training and inference, consequently they cannot use segment-level features. Such features are needed, for example, to identify figuration notes"
-
Symbolic chord recognition takes as input notes specified by onset, offset, pitch and metrical position, which can be extracted from formats such as MIDI, kern or MusicXML.
"we consider the music to be in symbolic form, i.e. as a collection of notes specified in terms of onset, offset, pitch, and metrical position. Symbolic representations can be extracted from formats such as MIDI, kern, or MusicXML."
-
Performance was lower on Mozart and Beethoven theme-and-variation sets than on Bach chorales, which the authors attribute to more figuration and rhythmic diversity in the variations.
"The comparatively lower performance on TAVERN is likely due to the substantially larger number of figurations and higher rhythmic diversity of the variations compared to the easier, mostly note-for-note texture of the chorales."
Available in