Continued Pretraining for Low-Resource Swahili ASR: Achieving State-of-the-Art Performance with Minimal Labeled Data
This work addresses the challenge of developing effective automatic speech recognition systems for low-resource languages, which is incremental but provides a replicable methodology for broader applications.
The paper tackled the problem of adapting speech recognition models to low-resource languages like Swahili by using continued pretraining with minimal labeled data, achieving a 3.24% word error rate, which is an 82% relative improvement over the baseline and surpasses prior academic systems by 61%.
We investigate continued pretraining (CPT) for adapting wav2vec2-bert-2.0 to Swahili automatic speech recognition (ASR). Our approach combines unlabeled audio with limited labeled data through pseudo-labeled CPT followed by supervised finetuning. With 20,000 labeled samples, we achieve 3.24% WER on Common Voice Swahili-an 82% relative improvement over the baseline. This result surpasses the best previously reported academic system (8.3% WER from XLS-R) by 61% relative improvement. We provide concrete data requirements and a replicable methodology applicable to other low-resource languages.