Speech models for Nigerian languages: progress and what the data tells us
We assembled 37.44 hours of speech across Hausa, Yoruba and Igbo, measured baselines, and ran fine-tuning experiments. The finding: public data plateaus — real consented telephony is what moves accuracy.
This is a progress report on Telroi’s speech-AI research — the datasets we have assembled, the baselines we have measured, and what our experiments tell us about building accurate speech recognition for Nigerian languages.
The bootstrap corpus
We assembled 37.44 hours of license-clean speech across 3 language tracks as a public-data starting point:
- Hausa — 13.62 h, 3259 clips (fleurs, CC-BY-4.0)
- Igbo — 13.79 h, 2839 clips (fleurs, CC-BY-4.0)
- Yoruba — 10.03 h, 2339 clips (fleurs, CC-BY-4.0)
Baselines: off-the-shelf accuracy
Word error rate (WER) of the off-the-shelf Meta MMS model on held-out read speech — the “before” we measure against:
- Hausa — 26.2% WER
- Yoruba — 55.9% WER
- Igbo — 44.6% WER
Experiment 1 — fine-tuning on public read speech
Fine-tuning on the clean public corpus produced almost no improvement. Results:
- Hausa: 26.2% → 26.3%
- Yoruba: 55.9% → 54.9%
- Igbo: 44.6% → 43.6%
The reason is instructive: the base model was already trained partly on this public data, so re-training on it is largely circular.
Experiment 2 — adapting to telephony conditions
We degraded the same speech to phone-call conditions (narrowband 8 kHz, codec quantization, noise) — audio the model has not seen — and adapted to it. Results:
- Hausa: 33.2% → 32.7% (telephony)
- Yoruba: 61% → 59.9% (telephony)
- Igbo: 60.3% → 58.9% (telephony)
Telephony conditions sharply degrade the off-the-shelf model — most of all for Igbo — confirming that real calls are far harder than clean recordings. Adapting on synthetic telephony recovered only a little, because the acoustic front-end that the degradation affects is not reached by lightweight adaptation, and synthetic audio is not real audio.
What this means
Across both experiments the conclusion is consistent and important: public data — clean or degraded — cannot by itself produce an accurate Nigerian-telephony speech model. The accuracy that matters comes only from real, consented telephony in the target languages. That is precisely the data Telroi is built to collect, with consent and de-identification enforced at the data layer, and it is the foundation the next phase of this work builds on.