CITATION — REFERENCE ENTRY
A Bi-Directional Transformer for Musical Chord Recognition — Proceedings of the 20th International Society for Music Information Retrieval Conference (ISMIR 2019)
- Key
- park2019-bidirectional-transformer
- Authors
- Park, Jonggwon; Choi, Kyoyun; Jeon, Sungwook; Kim, Dokyun; Park, Jonghun
- Issued
- 2019
- Type
- paper-conference
- Container
- Proceedings of the 20th International Society for Music Information Retrieval Conference (ISMIR 2019)
- Pages
- 620-627
Raw CSL JSON
{
"DOI": "10.5281/zenodo.3527886",
"URL": "https://archives.ismir.net/ismir2019/paper/000075.pdf",
"page": "620-627",
"type": "paper-conference",
"title": "A Bi-Directional Transformer for Musical Chord Recognition",
"author": [
{
"given": "Jonggwon",
"family": "Park"
},
{
"given": "Kyoyun",
"family": "Choi"
},
{
"given": "Sungwook",
"family": "Jeon"
},
{
"given": "Dokyun",
"family": "Kim"
},
{
"given": "Jonghun",
"family": "Park"
}
],
"issued": {
"date-parts": [
[
2019
]
]
},
"event-place": "Delft, Netherlands",
"container-title": "Proceedings of the 20th International Society for Music Information Retrieval Conference (ISMIR 2019)"
}
Claims
-
Attention-map analysis showed the model dividing its input into chord segments using the adaptive receptive field of the attention mechanism.
"It turns out that the model was able to divide segments of chords by utilizing adaptive receptive field of the attention mechanism."
-
A CRF decoder is trained mainly to smooth predicted chord sequences, which are often fragmented.
"The main purpose of training a CRF decoder is to smooth the predicted chord sequences that are often fragmented."
-
Park and colleagues identify Humphrey and Bello's convolutional neural network for major/minor chord classification as the first deep-learning chord recognition system.
"The very first deep-learning-based chord recognition system was proposed by [14] where they trained a CNN for"
-
The bi-directional transformer for chord recognition needs only a single training phase, without separate feature extractors or HMM/CRF decoders, and obtains results comparable to systems that use them.
"BTC requires only a single training phase while being able to obtain results comparable to them."
Available in