Mai Ho‘omāuna i ka ‘Ai: Language Models Improve Automatic Speech Recognition in Hawaiian

Kaavya D Chaparala, Guido Zarrella, Bruce Torres Fischer, Larry Kimura, Oiwi Parker Jones
Proceedings of The 4th NeurIPS Efficient Natural Language and Speech Processing Workshop, PMLR 262:576-583, 2024.

Abstract

In this paper we address the challenge of improving Automatic Speech Recognition (ASR) for a low-resource language, Hawaiian, by incorporating large amounts of independent text data into an ASR foundation model, Whisper. To do this, we train an external language model (LM) on ∼1.5M words of Hawaiian text. We then use the LM to rescore Whisper and compute word error rates (WERs) on a manually curated test set of labeled Hawaiian data. As a baseline, we use Whisper without an external LM. Experimental results reveal a small but significant improvement in WER when ASR outputs are rescored with a Hawaiian LM. The results support leveraging all available data in the development of ASR systems for underrepresented languages.

Cite this Paper

BibTeX
@InProceedings{pmlr-v262-d-chaparala24a, title = {{Mai Ho‘omāuna i ka ‘Ai}: Language Models Improve Automatic Speech Recognition in Hawaiian }, author = {D Chaparala, Kaavya and Zarrella, Guido and Torres Fischer, Bruce and Kimura, Larry and Parker Jones, Oiwi}, booktitle = {Proceedings of The 4th NeurIPS Efficient Natural Language and Speech Processing Workshop}, pages = {576--583}, year = {2024}, editor = {Rezagholizadeh, Mehdi and Passban, Peyman and Samiee, Soheila and Partovi Nia, Vahid and Cheng, Yu and Deng, Yue and Liu, Qun and Chen, Boxing}, volume = {262}, series = {Proceedings of Machine Learning Research}, month = {14 Dec}, publisher = {PMLR}, pdf = {https://raw.githubusercontent.com/mlresearch/v262/main/assets/d-chaparala24a/d-chaparala24a.pdf}, url = {https://proceedings.mlr.press/v262/d-chaparala24a.html}, abstract = {In this paper we address the challenge of improving Automatic Speech Recognition (ASR) for a low-resource language, Hawaiian, by incorporating large amounts of independent text data into an ASR foundation model, Whisper. To do this, we train an external language model (LM) on ∼1.5M words of Hawaiian text. We then use the LM to rescore Whisper and compute word error rates (WERs) on a manually curated test set of labeled Hawaiian data. As a baseline, we use Whisper without an external LM. Experimental results reveal a small but significant improvement in WER when ASR outputs are rescored with a Hawaiian LM. The results support leveraging all available data in the development of ASR systems for underrepresented languages.} }
Endnote
%0 Conference Paper %T Mai Ho‘omāuna i ka ‘Ai: Language Models Improve Automatic Speech Recognition in Hawaiian %A Kaavya D Chaparala %A Guido Zarrella %A Bruce Torres Fischer %A Larry Kimura %A Oiwi Parker Jones %B Proceedings of The 4th NeurIPS Efficient Natural Language and Speech Processing Workshop %C Proceedings of Machine Learning Research %D 2024 %E Mehdi Rezagholizadeh %E Peyman Passban %E Soheila Samiee %E Vahid Partovi Nia %E Yu Cheng %E Yue Deng %E Qun Liu %E Boxing Chen %F pmlr-v262-d-chaparala24a %I PMLR %P 576--583 %U https://proceedings.mlr.press/v262/d-chaparala24a.html %V 262 %X In this paper we address the challenge of improving Automatic Speech Recognition (ASR) for a low-resource language, Hawaiian, by incorporating large amounts of independent text data into an ASR foundation model, Whisper. To do this, we train an external language model (LM) on ∼1.5M words of Hawaiian text. We then use the LM to rescore Whisper and compute word error rates (WERs) on a manually curated test set of labeled Hawaiian data. As a baseline, we use Whisper without an external LM. Experimental results reveal a small but significant improvement in WER when ASR outputs are rescored with a Hawaiian LM. The results support leveraging all available data in the development of ASR systems for underrepresented languages.
APA
D Chaparala, K., Zarrella, G., Torres Fischer, B., Kimura, L. & Parker Jones, O.. (2024). Mai Ho‘omāuna i ka ‘Ai: Language Models Improve Automatic Speech Recognition in Hawaiian . Proceedings of The 4th NeurIPS Efficient Natural Language and Speech Processing Workshop, in Proceedings of Machine Learning Research 262:576-583 Available from https://proceedings.mlr.press/v262/d-chaparala24a.html.

Related Material