EUVIMEDEUVIMED entdecken
Crossref

Using articulatory feature detectors in progressive networks for multilingual low-resource phone recognition

Mahir Morshed, Mark Hasegawa-Johnson iD

The Journal of the Acoustical Society of America · 2024

DOI: 10.1121/10.0034415 ↗

Kurz gefasst

Fachartikel in der Zeitschrift The Journal of the Acoustical Society of America, 2024. Verfasst von Mahir Morshed und 1 weiteren Autor:innen. Beteiligte Einrichtung: Department of Electrical and Computer Engineering, University of Illinois Urbana-Champaign, Urbana, Illinois 61801, USA. Hauptthemen laut MEDLINE: Phonetics, Neural Networks, Computer, Multilingualism. Die Arbeit nennt 35 Literaturangaben.

AbstractSchlagwörterBibliografische AngabenZitierenVerwandte Studien

Abstract (Originalsprache)

Systems inspired by progressive neural networks, transferring information from end-to-end articulatory feature detectors to similarly structured phone recognizers, are described. These networks, connecting the corresponding recurrent layers of pre-trained feature detector stacks and newly introduced phone recognizer stacks, were trained on data from four Asian languages, with experiments testing the system on those languages and four African languages. Later adjustments of these networks include the use of contrastive predictive coding layers at the inputs to those networks' recurrent portions. Such adjustments allow for performance differences to be attributed to the presence or absence of individual feature detectors (for consonant place/manner and vowel height/backness). Some of these differences manifest after feature-level comparisons of recognizer outputs, as well as through considering variations and ablations in architecture and training setup. These differences encourage further exploration of methods to reduce errors with phones having specific articulatory features as well as further architectural modifications.

Schlagwörter

MeSH-Schlagwörter (MEDLINE)

PhoneticsNeural Networks, ComputerMultilingualismHumansSpeech AcousticsSpeech Recognition SoftwareSpeech Production MeasurementLanguage

Autor:innen und Institutionen

  • Department of Electrical and Computer Engineering, University of Illinois Urbana-Champaign, Urbana, Illinois 61801, USA.

ORCID-Profile: Mark Hasegawa-Johnson

Zitieren

APA

Morshed, M., & Hasegawa-Johnson, M. (2024). Using articulatory feature detectors in progressive networks for multilingual low-resource phone recognition. The Journal of the Acoustical Society of America. https://doi.org/10.1121/10.0034415

Vancouver

Morshed M, Hasegawa-Johnson M. Using articulatory feature detectors in progressive networks for multilingual low-resource phone recognition. The Journal of the Acoustical Society of America. 2024. doi:10.1121/10.0034415

BibTeX
@article{Morshed2024,
  author = {Morshed, Mahir and Hasegawa-Johnson, Mark},
  title = {Using articulatory feature detectors in progressive networks for multilingual low-resource phone recognition},
  journal = {The Journal of the Acoustical Society of America},
  year = {2024},
  doi = {10.1121/10.0034415}
}
Kostenlos starten →