Mai Ho'omāuna i ka 'Ai: Language Models Improve Automatic Speech Recognition in Hawaiian
This work addresses the problem of enhancing ASR for underrepresented languages like Hawaiian, though it is incremental as it applies a standard LM rescoring technique to a new language.
The paper tackled the challenge of improving Automatic Speech Recognition (ASR) for the low-resource language Hawaiian by training an external language model on ~1.5M words of Hawaiian text and using it to rescore outputs from the Whisper foundation model, resulting in a small but significant improvement in word error rates (WERs) on a manually curated test set.
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.