Skip to Main content Skip to Navigation
Conference papers

Spécialisation de modèles neuronaux pour la transcription phonémique : premiers pas vers la reconnaissance de mots pour les langues rares

Abstract : We describe the latest results we have obtained in the development of NLP (Natural Language Processing) tools to reduce the transcription and annotation workload of field linguists, as part of workflows to document and describe the world's languages. We show how a new deep learning approach based on the fine-tuning of a generic representation model allows to significantly improve the quality of automatic phonemic transcription, and, more significantly, to take a first step towards automatic word recognition for low-resource languages.
Document type :
Conference papers
Complete list of metadata

https://halshs.archives-ouvertes.fr/halshs-03475443
Contributor : Alexis Michaud Connect in order to contact the contributor
Submitted on : Friday, December 10, 2021 - 9:56:27 PM
Last modification on : Tuesday, January 4, 2022 - 5:51:56 AM

Licence


Distributed under a Creative Commons Attribution - NonCommercial - ShareAlike 4.0 International License

Identifiers

  • HAL Id : halshs-03475443, version 1

Citation

Cécile Macaire, Guillaume Wisniewski, Séverine Guillaume, Benjamin Galliot, Guillaume Jacques, et al.. Spécialisation de modèles neuronaux pour la transcription phonémique : premiers pas vers la reconnaissance de mots pour les langues rares. Journées scientifiques du Groupement de recherche "Linguistique informatique, formelle et de terrain" (GDR LIFT), Dec 2021, Grenoble, France. ⟨halshs-03475443⟩

Share

Metrics

Les métriques sont temporairement indisponibles