Whisper Large V3 — Swahili NSS (AfriVoice 5 Domains)
Stage 2 nonstandard speech fine-tune of openai/whisper-large-v3 on the CDLI Kenyan Swahili Nonstandard Speech dataset. Built on top of ElizabethMwangi/whisper-large-v3-swahili-afrivoice-5domains, a Stage 1 checkpoint trained on AfriVoice Swahili across all 5 domains (389,599 utterances), best step 600.
Training Details
- Stage 1 base: ElizabethMwangi/whisper-large-v3-swahili-afrivoice-5domains (AfriVoice 5 domains, 389,599 utterances)
- Stage 2 dataset: cdli/kenyan_swahili_nonstandard_speech_v1.0 (849 examples)
- Language token: sw
- Full model training: encoder, decoder, and projection updated
- SpecAugment: On
- Learning rate: 5e-5 (polynomial decay, warmup 40 steps)
- Max steps: 1000
- Best checkpoint: step 650
- Output dir: sw_nss_tune_whisper_v3_large_afrivoice_v1.3
Results on Kenyan Swahili Nonstandard Speech (test set, best checkpoint)
Overall, compared to unadapted baseline
| Model | Overall WER | Avg WER |
|---|---|---|
| openai/whisper-large-v3 (no adaptation) | 0.845 | - |
| Stage 1 only (afrivoice-5domains, no NSS tuning) | 0.434 | - |
| This model (adapted) | 0.329 | 0.279 |
| Metric | Value |
|---|---|
| Overall WER | 0.329 |
| Overall CER | 0.171 |
| Avg WER | 0.279 |
| Avg CER | 0.128 |
| Meaning preservation | 75.7% |
By severity
| Severity | WER | CER |
|---|---|---|
| Mild | 0.21 | 0.09 |
| Moderate | 0.25 | 0.11 |
| Severe | 0.35 | 0.16 |