Google Crowdsourced Speech Corpora and Related Open-Source Resources for Low-Resource Languages and Dialects: An Overview

Butryna, Alena; Chu, Shan-Hui Cathy; Demirsahin, Isin; Gutkin, Alexander; Ha, Linne; He, Fei; Jansche, Martin; Johny, Cibu C; Anna, Katanova,; Kjartansson, Oddur; Li, Chenfang; Merkulova, T. I.; Oo, Yin May; Pipatsrisawat, Knot; Rivera, Clara; Sarin, Supheakmungkol; Silva, Pasindu de; Sodimana, Keshan; Sproat, Richard; Wattanavekin, Theeraphol; Wibawa, Jaka Aris Eko

doi:10.48550/arxiv.2010.06778

Cited by 2 publications

(3 citation statements)

References 0 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…The NCHLT project expands the efforts of the works in Roux et al ( 2004 ) and Badenhorst et al ( 2011 ) to enable the development of large-vocabulary speech recognition systems for practical applications. There are other existing databases, such as the Wall Street Journal (Garofolo, Graff, Paul, & Pallett, 1993 ), GlobalPhone (Schultz, 2002 ) and Google (Butryna et al, 2020 ) corpora. However, these only contain data for the English language.…”

Section: Resultsmentioning

confidence: 99%

A proposed artificial intelligence-based real-time speech-to-text to sign language translator for South African official languages for the COVID-19 era and beyond: In pursuit of solutions for the hearing impaired

Madahana

Khoza-Shangase

Moroe

et al. 2022

SAJCD

View full text Add to dashboard Cite

Background: The emergence of the coronavirus disease 2019 (COVID-19) pandemic has resulted in communication being heightened as one of the critical aspects in the implementation of interventions. Delays in the relaying of vital information by policymakers have the potential to be detrimental, especially for the hearing impaired.Objectives: This study aims to conduct a scoping review on the application of artificial intelligence (AI) for real-time speech-to-text to sign language translation and consequently propose an AI-based real-time translation solution for South African languages from speech-to-text to sign language.Methods: Electronic bibliographic databases including ScienceDirect, PubMed, Scopus, MEDLINE and ProQuest were searched to identify peer-reviewed publications published in English between 2019 and 2021 that provided evidence on AI-based real-time speech-to-text to sign language translation as a solution for the hearing impaired. This review was done as a precursor to the proposed real-time South African translator.Results: The review revealed a dearth of evidence on the adoption and/or maximisation of AI and machine learning (ML) as possible solutions for the hearing impaired. There is a clear lag in clinical utilisation and investigation of these technological advances, particularly in the African continent.Conclusion: Assistive technology that caters specifically for the South African community is essential to ensuring a two-way communication between individuals who can hear clearly and individuals with hearing impairments, thus the proposed solution presented in this article.

show abstract

Section: Resultsmentioning

confidence: 99%

A proposed artificial intelligence-based real-time speech-to-text to sign language translator for South African official languages for the COVID-19 era and beyond: In pursuit of solutions for the hearing impaired

Madahana

Khoza-Shangase

Moroe

et al. 2022

SAJCD

View full text Add to dashboard Cite

show abstract

“…The dataset will be used from (Butryna et al, 2020), which contains compilations of speech audio with .wav extension, file ID, and transcription of each audio file. There is a total of 5822 audio files with their ID and transcription, with an average duration of each audio being 3 seconds, and the total duration of all audio files is plus minus 4.85 hours.…”

Section: Data Collectionmentioning

confidence: 99%

“…Low-resource language is a language that has little data in the form of digital, or that can be processed by a computer directly. One example of this language category is the Javanese language (Butryna et al, 2020). Whisper performance evaluated using Character Error Rate (CER)/Word Error Rate (WER) on this type of language is remarkably low.…”

Section: Introductionmentioning

confidence: 99%

Analysis of Whisper Automatic Speech Recognition Performance on Low Resource Language

Pratama,

Amrullah

2024

pilar

View full text Add to dashboard Cite

Implementing Automatic Speech Recognition Technology in daily life could give convenience to its users. However, speeches that can be recognized accurately by the ASR model right now are in languages considered high resources, like English. In previous research, a few regional languages like Javanese, Sundanese, Balinese and Btaknese are used in automatic speech recognition. This research aim is to improve speech recognition using the ASR model on low-resource language. The dataset used in this research is the Javanese dataset specifically because there is a high-quality Javanese speech dataset provided by previous research. The method used is fine-tuning the Whisper model which has been trained on 680,000 hours of multilingual voice data using a Javanese speech dataset. To reduce computation requirements, parameter efficient fine-tuning (PEFT) implemented in the fine-tuning process. The trainable parameter is reduced to <1% because the implementation of PEFT reduces the computation required by the model for fine-tuning. The best WER evaluation result is 13.77%, achieved by the fine-tuned Whisper large-v2 model compared to the base model of Whisper large-v2, which achieves 89.40% in WER evaluation. Performance improvement in WER evaluation showed that fine-tuning effectively improves the performance of the Whisper automatic speech recognition model on recognizing speeches in low-resource languages like the Javanese language compared to the Original Whisper model performance with minimal computational cost needed for fine-tuning large model.

show abstract

Google Crowdsourced Speech Corpora and Related Open-Source Resources for Low-Resource Languages and Dialects: An Overview

Cited by 2 publications

References 0 publications

A proposed artificial intelligence-based real-time speech-to-text to sign language translator for South African official languages for the COVID-19 era and beyond: In pursuit of solutions for the hearing impaired

A proposed artificial intelligence-based real-time speech-to-text to sign language translator for South African official languages for the COVID-19 era and beyond: In pursuit of solutions for the hearing impaired

Analysis of Whisper Automatic Speech Recognition Performance on Low Resource Language

Contact Info

Product

Resources

About