Finetuned openai/whisper-small on 12954 Swedish training audio samples from artifacts/sv-SE_mozilla-foundation_common_voice_17_0.
This model was created from the Mozilla.ai Blueprint: speech-to-text-finetune.
Evaluation results on 5259 audio samples of Swedish:
Baseline model (before finetuning) on Swedish
- Word Error Rate: 28.406
- Loss: 1.068
Finetuned model (after finetuning) on Swedish
- Word Error Rate: 20.889
- Loss: 0.305