# Tone-Conditioned Curriculum Learning for Low-Resource Bantu Speech Recognition

arXiv: https://arxiv.org/abs/2606.31642 · HTML: https://mlsanity.com/paper/2606.31642/

## Key facts

- No citing papers in the map yet (data through 14 Oct 2026).
- Builds on 8 of its 30 references that are in the map.
- First posted on arXiv 30 Jun 2026.

## Builds on (most cited first, 8 of 8)

- [Robust Speech Recognition via Large-Scale Weak Supervision](https://mlsanity.com/paper/2212.04356/) (2022)
- [Scaling Speech Technology to 1,000+ Languages](https://mlsanity.com/paper/2305.13516/) (2023)
- [Curriculum Pre-training for End-to-End Speech Translation](https://mlsanity.com/paper/2004.10093/) (2020)
- [Curriculum-based transfer learning for an effective end-to-end spoken language understanding and domain portability](https://mlsanity.com/paper/1906.07601/) (2019)
- [Multilingual self-supervised speech representations improve the speech recognition of low-resource African languages with codeswitching](https://mlsanity.com/paper/2311.15077/) (2023)
- [Swivuriso: The South African Next Voices Multilingual Speech Dataset](https://mlsanity.com/paper/2512.02201/) (2025)
- [Comparison and Analysis of New Curriculum Criteria for End-to-End ASR](https://mlsanity.com/paper/2208.05782/) (2022)
- [KinSPEAK: Improving speech recognition for Kinyarwanda via semi-supervised learning methods](https://mlsanity.com/paper/2308.11863/) (2023)

## Abstract (arXiv)

Southern Bantu languages are spoken by over 80 million people, yet current foundation ASR models still produce zero-shot WER above 100%, which limits practical use in education and public services. We addressed this gap with a tone conditioned curriculum framework for 6 Southern Bantu languages that combined hybrid difficulty scoring, gated adapters driven by tonal statistics and staged curriculum training. We trained on a community corpus and tested transfer to NCHLT to measure robustness beyond matched evaluation. Results revealed clear interactions between architecture and language, with W2V-BERT outperforming Whisper on Nguni languages by 3 to 4 WER points whilst Whisper performed better on Sotho-Tswana languages. W2V-BERT with tone conditioning reached 28.41% average WER across datasets and 23.79% on Xitsonga transfer. No single model suited all 6 languages, so deployment should pair model selection per language with validation across corpora.