CTC-CRF 4gram-LM

Reported on 5 benchmarks across 1 task · 1 paper · 1 SOTA

Note: results are matched by exact model name. Different papers may use the same name for different model variants.

Audio5 results

Speech RecognitiononAISHELL-1
Word Error Rate (WER)· 2020-05-27
6.34
best: 0.55 (FireRedASR-AED)
SOTA
CAT: A CTC-CRF based ASR Toolkit Bridging the Hybrid and the End-to-end Approaches towards Data Efficiency and Low Latency arXiv:2005.13326
Speech RecognitiononWSJ eval92
Word Error Rate (WER)
3.79
best: 1.3 (Speechstew 100M)
Speech RecognitiononWSJ eval93
Word Error Rate (WER)
6.23
best: 4.98 (Deep Speech 2)
Speech RecognitiononLibriSpeech test-clean
Word Error Rate (WER)
4.09
best: 0.985 (United Med ASR)
Speech RecognitiononLibriSpeech test-other
Word Error Rate (WER)
10.65
best: 2.48 (SAMBA ASR)