LEARNING UNSUPERVISED HIERARCHIES OF AUDIO CONCEPTS - Archive ouverte HAL Accéder directement au contenu
Communication Dans Un Congrès Année : 2022

LEARNING UNSUPERVISED HIERARCHIES OF AUDIO CONCEPTS

Résumé

Music signals are difficult to interpret from their lowlevel features, perhaps even more than images: e.g. highlighting part of a spectrogram or an image is often insufficient to convey high-level ideas that are genuinely relevant to humans. In computer vision, concept learning was therein proposed to adjust explanations to the right abstraction level (e.g. detect clinical concepts from radiographs). These methods have yet to be used for MIR. In this paper, we adapt concept learning to the realm of music, with its particularities. For instance, music concepts are typically non-independent and of mixed nature (e.g. genre, instruments, mood), unlike previous work that assumed disentangled concepts. We propose a method to learn numerous music concepts from audio and then automatically hierarchise them to expose their mutual relationships. We conduct experiments on datasets of playlists from a music streaming service, serving as a few annotated examples for diverse concepts. Evaluations show that the mined hierarchies are aligned with both ground-truth hierarchies of concepts ± when available ± and with proxy sources of concept similarity in the general case.
Fichier principal
Vignette du fichier
000051.pdf (982.82 Ko) Télécharger le fichier
Origine : Fichiers éditeurs autorisés sur une archive ouverte

Dates et versions

hal-03909873 , version 1 (21-12-2022)

Identifiants

  • HAL Id : hal-03909873 , version 1

Citer

Darius Afchar, Romain Hennequin, Vincent Guigue. LEARNING UNSUPERVISED HIERARCHIES OF AUDIO CONCEPTS. 23rd International Society for Music Information Retrieval Conference (ISMIR 2022), Dec 2022, Bengaluru,, India. ⟨hal-03909873⟩
6 Consultations
19 Téléchargements

Partager

Gmail Facebook X LinkedIn More