CITATION.cff

cff-version: 1.2.0
message: "If you use this work in a project of yours and write about it, please cite our ACL 2022 paper using the following citation data."
title: "Adapted FastSpeech2"
version: 1.0.0
url: "https://github.com/roedoejet/FastSpeech2"
preferred-citation:
  type: conference-paper
  title: >-
    Requirements and Motivations of Low-Resource Speech Synthesis for Language Revitalization
  authors:
    - given-names: Aidan
      family-names: Pine
      email: Aidan.Pine@nrc-cnrc.gc.ca
      affiliation: National Research Council Canada
    - given-names: Dan
      family-names: Wells
      email: dan.wells@ed.ac.uk
      affiliation: University of Edinburgh
    - family-names: "Nathan Thanyehténhas"
      given-names: "Brinklow"
      email: Patrick.Littell@nrc-cnrc.gc.ca
      affiliation: Queen's University
    - given-names: Patrick
      family-names: Littell
      email: Patrick.Littell@nrc-cnrc.gc.ca
      affiliation: National Research Council Canada
    - given-names: Korin
      family-names: Richmond
      email: korin.richmond@ed.ac.uk
      affiliation: University of Edinburgh
  collection-title: >-
    "Proceedings of the 60th Annual Meeting of the Association for Computational Linguistics"
  year: 2022
  month: 5
  publisher:
    name: Association for Computational Linguistics
  url: https://aclanthology.org/2022.acl-long.507/
  start: 7346
  end: 7359
  location:
    name: Dublin, Ireland
  abstract: >-
    This paper describes the motivation and development of speech synthesis systems for the purposes of language revitalization. By building speech synthesis systems for three Indigenous languages spoken in Canada, Kanien’kéha, Gitksan & SENĆOŦEN, we re-evaluate the question of how much data is required to build low-resource speech synthesis systems featuring state-of-the-art neural models. For example, preliminary results with English data show that a FastSpeech2 model trained with 1 hour of training data can produce speech with comparable naturalness to a Tacotron2 model trained with 10 hours of data. Finally, we motivate future research in evaluation and classroom integration in the field of speech synthesis for language revitalization.