Speech Recognition On Common Voice French
Métriques
Test WER
Résultats
Résultats de performance de divers modèles sur ce benchmark
Nom du modèle | Test WER | Paper Title | Repository |
|---|---|---|---|
| QuartzNet15x5FR (CV-only) | 12.1% | Scribosermo: Fast Speech-to-Text models for German and other Languages | |
| ConformerCTC-L (5-gram) | 8.13% | Scribosermo: Fast Speech-to-Text models for German and other Languages | |
| ConformerCTC-L (4-gram) | 9.16% | NeMo: a toolkit for building AI applications using Neural Modules | |
| ConformerCTC-L (no-LM) | 10.19 % | Scribosermo: Fast Speech-to-Text models for German and other Languages | |
| VoxPopuli-50K (n-gram) | 9.6% | VoxPopuli: A Large-Scale Multilingual Speech Corpus for Representation Learning, Semi-Supervised Learning and Interpretation | |
| ConformerCTC-L (no-LM) | 9.63% | NeMo: a toolkit for building AI applications using Neural Modules | |
| Whisper (Large v2) | 13.9% | Robust Speech Recognition via Large-Scale Weak Supervision | |
| QuartzNet15x5FR (D7) | 11.0% | Scribosermo: Fast Speech-to-Text models for German and other Languages |
0 of 8 row(s) selected.