Speech Recognition On Tuda
評価指標
Test WER
評価結果
このベンチマークにおける各モデルのパフォーマンス結果
モデル名 | Test WER | Paper Title | Repository |
|---|---|---|---|
| QuartzNet15x5DE (D37) | 10.2% | Scribosermo: Fast Speech-to-Text models for German and other Languages | |
| Kaldi | 20.5% | Open Source German Distant Speech Recognition: Corpus and Acoustic Model | - |
| Conformer-Transducer (no LM) | 5.82% | Automatic Speech Recognition in German: A Detailed Error Analysis | - |
| DeepSpeech-Polyglot | 18.6% | - | - |
| Kaldi | 14.4% | Open Source Automatic Speech Recognition for German | |
| IMS-Speech | 12.0% | IMS-Speech: A Speech to Text Tool | - |
| Hybrid CTC/Attention | 12.8% | CTC-Segmentation of Large Corpora for German End-to-end Speech Recognition | |
| PocketSphinx | 39.6% | Open Source German Distant Speech Recognition: Corpus and Acoustic Model | - |
| TDNN-HMM hybrid, FST (with RNNLM rescoring) | 6.93% | - | - |
0 of 9 row(s) selected.