SubmissionNumber#=%=#21
FinalPaperTitle#=%=#Child Support: Leveraging Lexifiers Resources to Support Creoles ASR
ShortPaperTitle#=%=#
NumberOfPages#=%=#7
CopyrightSigned#=%=#Eric Le Ferrand
JobTitle#==#
Organization#==#
Abstract#==#Creole languages emerged from colonial contact and the slave trade. Although they inherit
the bulk of their vocabulary from a "lexifier"
language, they remain classic low-resource
languages, presenting significant challenges
for speech technology. This paper explores
how the abundant resources of a lexifier can
be leveraged for Creole-specific tools, focusing on Automatic Speech Recognition (ASR).
Specifically, we use an artificial dataset generated a French-trained Text-to-Speech (TTS)
model and French datasets to pre-finetune ASR
models for two French-based Creoles. Our
results demonstrate that a two-stage training
setup where models are first trained on artificial datasets leads to substantial performance
boost for transcribing Creole languages. Additionally, this approach serves as a viable first
step for ASR development in zero-resource scenarios.
Author{1}{Firstname}#=%=#Éric
Author{1}{Lastname}#=%=#Le Ferrand
Author{1}{Username}#=%=#leferrae
Author{1}{Orcid}#=%=#
Author{1}{Email}#=%=#leferran@bc.edu
Author{1}{Affiliation}#=%=#Boston College
Author{2}{Firstname}#=%=#Fabiola
Author{2}{Lastname}#=%=#Henri
Author{2}{Orcid}#=%=#
Author{2}{Email}#=%=#fabiolah@buffalo.edu
Author{2}{Affiliation}#=%=#University at Buffalo

==========
èéáğö