Fine-tuning pre-trained models for Automatic Speech Recognition: experiments on a fieldwork corpus of Japhug (Trans-Himalayan family)
Fiche du document
26 mai 2022
- handle: 10670/1.f4q871
- halshs: halshs-03647315
- doi: 10.18653/v1/2022.computel-1.21
Ce document est lié à :
info:eu-repo/semantics/altIdentifier/doi/10.18653/v1/2022.computel-1.21
http://creativecommons.org/licenses/by-nc-sa/ , info:eu-repo/semantics/OpenAccess
Mots-clés
Automatic Speech RecognitionSujets proches
Speech--Transcription Language and languages--Transcription Transcribing Learning process Family Family structure Families--Social aspects Families--Social conditions Family life Structure, Family Family relationships Relationships, Family Mechanical speech recognizer Speech recognition, Automatic Scholarship Learning Transcribing Biography--Ancestry Relations with family Biography--Descendants Biography--Family Ancestry Field work Field study Field protocolsCiter ce document
Séverine Guillaume et al., « Fine-tuning pre-trained models for Automatic Speech Recognition: experiments on a fieldwork corpus of Japhug (Trans-Himalayan family) », HAL-SHS : linguistique, ID : 10.18653/v1/2022.computel-1.21
Métriques
Partage / Export
Résumé
This is a report on results obtained in the development of speech recognition tools intended to support linguistic documentation efforts. The test case is an extensive fieldwork corpus of Japhug, an endangered language of the Trans-Himalayan (Sino-Tibetan) family. The goal is to reduce the transcription workload of field linguists. The method used is a deep learning approach based on the language-specific tuning of a generic pre-trained representation model, XLS-R, using a Transformer architecture. We note difficulties in implementation, in terms of learning stability. But this approach brings significant improvements nonetheless. The quality of phonemic transcription is improved over earlier experiments; and most significantly, the new approach allows for reaching the stage of automatic word recognition. Subjective evaluation of the tool by the author of the training data confirms the usefulness of this approach.