Speech Recognition On Common Voice Spanish
Metriken
Test WER
Ergebnisse
Leistungsergebnisse verschiedener Modelle zu diesem Benchmark
Modellname | Test WER | Paper Title | Repository |
|---|---|---|---|
| ConformerCTC-L (4-gram) | 5.5% | NeMo: a toolkit for building AI applications using Neural Modules | |
| QuartzNet15x5ES (CV-only) | 10.5% | Scribosermo: Fast Speech-to-Text models for German and other Languages | |
| ConformerCTC-L (5-gram) | 5.68% | Scribosermo: Fast Speech-to-Text models for German and other Languages | |
| ConformerCTC-L (no LM) | 6.9% | NeMo: a toolkit for building AI applications using Neural Modules | |
| VoxPopuli-50K (n-gram) | 10.0% | VoxPopuli: A Large-Scale Multilingual Speech Corpus for Representation Learning, Semi-Supervised Learning and Interpretation | |
| ConformerCTC-L (no-LM) | 7.46 % | Scribosermo: Fast Speech-to-Text models for German and other Languages | |
| Whisper (Large v2) | 5.6% | Robust Speech Recognition via Large-Scale Weak Supervision | |
| QuartzNet15x5ES (D8) | 10.0% | Scribosermo: Fast Speech-to-Text models for German and other Languages |
0 of 8 row(s) selected.