Speech Sequence Embeddings using Nearest Neighbors Contrastive Learning - Inria - Institut national de recherche en sciences et technologies du numérique Accéder directement au contenu
Communication Dans Un Congrès Année : 2022

Speech Sequence Embeddings using Nearest Neighbors Contrastive Learning

Résumé

We introduce a simple neural encoder architecture that can be trained using an unsupervised contrastive learning objective which gets its positive samples from data-augmented k-Nearest Neighbors search. We show that when built on top of recent self-supervised audio representations [1, 2, 3], this method can be applied iteratively and yield competitive SSE as evaluated on two tasks: query-by-example of random sequences of speech, and spoken term discovery. On both tasks our method pushes the state-of-the-art by a significant margin across 5 different languages. Finally, we establish a benchmark on a query-byexample task on the LibriSpeech dataset to monitor future improvements in the field.
Fichier principal
Vignette du fichier
2204.05148.pdf (520.05 Ko) Télécharger le fichier
Origine Fichiers produits par l'(les) auteur(s)

Dates et versions

hal-03831888 , version 1 (27-10-2022)

Identifiants

Citer

Robin Algayres, Adel Nabli, Benoît Sagot, Emmanuel Dupoux. Speech Sequence Embeddings using Nearest Neighbors Contrastive Learning. Interspeech 2022 - 23rd INTERSPEECH Conference, Sep 2022, Incheon, South Korea. ⟨hal-03831888⟩
36 Consultations
74 Téléchargements

Altmetric

Partager

Gmail Mastodon Facebook X LinkedIn More