Continued Pretraining for Low-Resource Swahili ASR: Achieving State-of-the-Art Performance with Minimal Labeled Data

TL;DR

This paper demonstrates that continued pretraining with pseudo-labeling significantly improves Swahili ASR performance using minimal labeled data, achieving state-of-the-art results.

Contribution

It introduces a combined pseudo-labeled CPT and supervised finetuning approach for low-resource Swahili ASR, surpassing previous benchmarks.

Findings

Achieved 3.24% WER with 20,000 labeled samples

Surpassed previous best by 61% relative improvement

Provides a replicable methodology for low-resource languages

Abstract

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.