Synthetic Voice Data for Automatic Speech Recognition in African Languages

TL;DR

This paper systematically assesses the use of large-scale synthetic voice data generated via LLM-driven text creation and TTS for improving African language ASR, demonstrating cost-effective performance gains and highlighting the need for better evaluation protocols.

Contribution

It introduces a novel three-step process for creating synthetic African language speech data and evaluates its effectiveness in improving ASR performance across multiple low-resource languages.

Findings

Synthetic data achieved high readability scores in most languages.

ASR models trained on synthetic data matched or exceeded real-data baselines.

Cost-effective synthetic data generation significantly improved low-resource language ASR.

Abstract

Speech technology remains out of reach for most of the over 2300 languages in Africa. We present the first systematic assessment of large-scale synthetic voice corpora for African ASR. We apply a three-step process: LLM-driven text creation, TTS voice synthesis, and ASR fine-tuning. Eight out of ten languages for which we create synthetic text achieved readability scores above 5 out of 7. We evaluated ASR improvement for three (Hausa, Dholuo, Chichewa) and created more than 2,500 hours of synthetic voice data at below 1% of the cost of real data. Fine-tuned Wav2Vec-BERT-2.0 models trained on 250h real and 250h synthetic Hausa matched a 500h real-data-only baseline, while 579h real and 450h to 993h synthetic data created the best performance. We also present gender-disaggregated ASR performance evaluation. For very low-resource languages, gains varied: Chichewa WER improved about 6.5% relative with a 1:2 real-to-synthetic ratio; a 1:1 ratio for Dholuo showed similar improvements on some evaluation data, but not on others. Investigating intercoder reliability, ASR errors and evaluation datasets revealed the need for more robust reviewer protocols and more accurate evaluation data. All data and models are publicly released to invite further work to improve synthetic data for African languages.