Skip to Main content Skip to Navigation
Conference papers

Autoapprentissage pour le regroupement en locuteurs : premières investigations

Abstract : This paper investigates self trained cross-show speaker diarization applied to collections of French TV archives, based on an i-vector/PLDA framework. The parameters used for i-vectors extraction and PLDA scoring are trained in a unsupervised way, using the data of the collection itself. Performances are compared, using combinations of target data and external data for training. The experimental results on two distinct target corpora show that using data from the corpora themselves to perform unsupervised iterative training and domain adaptation of PLDA parameters can improve an existing system, trained on external annotated data. Such results indicate that performing speaker indexation on small collections of unlabeled audio archives should only rely on the availability of a sufficient external corpus, which can be specifically adapted to every target collection. We show that a minimum collection size is required to exclude the use of such an external bootstrap.
Complete list of metadatas

https://hal.archives-ouvertes.fr/hal-01433156
Contributor : Sylvain Meignier <>
Submitted on : Tuesday, March 21, 2017 - 11:57:59 PM
Last modification on : Wednesday, March 4, 2020 - 4:15:21 PM
Document(s) archivé(s) le : Thursday, June 22, 2017 - 2:54:19 PM

File

V01-JEP - copie.pdf
Publisher files allowed on an open archive

Identifiers

  • HAL Id : hal-01433156, version 1

Collections

Citation

Gaël Le Lan, Sylvain Meignier, Delphine Charlet, Anthony Larcher. Autoapprentissage pour le regroupement en locuteurs : premières investigations. Journées d’Études sur la Parole (JEP'16), 2016, Paris, France. pp.80-82. ⟨hal-01433156⟩

Share

Metrics

Record views

131

Files downloads

47