Impact of time and note duration tokenizations on deep learning symbolic music modeling - Archive ouverte HAL Access content directly
Conference Papers Year : 2023

Impact of time and note duration tokenizations on deep learning symbolic music modeling

Nathan Fradet
  • Function : Author
  • PersonId : 1116107
Fabien Chhel
Jean-Pierre Briot

Abstract

Symbolic music is widely used in various deep learning tasks, including generation, transcription, synthesis, and Music Information Retrieval (MIR). It is mostly employed with discrete models like Transformers, which require music to be tokenized, i.e., formatted into sequences of distinct elements called tokens. Tokenization can be performed in different ways, and recent research has focused on developing more efficient methods. However, the key differences between these methods are often unclear, and few studies have compared them. In this work, we analyze the current common tokenization methods and experiment with time and note duration representations. We compare the performance of these two impactful criteria on several tasks, including composer classification, emotion classification, music generation, and sequence representation. We demonstrate that explicit information leads to better results depending on the task.
Fichier principal
Vignette du fichier
ISMIR_2023.pdf (549.2 Ko) Télécharger le fichier
Origin : Files produced by the author(s)

Dates and versions

hal-04147659 , version 1 (26-10-2023)

Licence

Copyright

Identifiers

Cite

Nathan Fradet, Nicolas Gutowski, Fabien Chhel, Jean-Pierre Briot. Impact of time and note duration tokenizations on deep learning symbolic music modeling. 24th Conference of the International Society for Music Information Retrieval (ISMIR) 2023, Augusto Sarti; Fabio Antonacci; Mark Sandler, Nov 2023, Milano, Italy. pp.89-97, ⟨10.5281/zenodo.10265229⟩. ⟨hal-04147659⟩
99 View
8 Download

Altmetric

Share

Gmail Facebook X LinkedIn More