RNN-T

Reported on 4 benchmarks across 4 tasks · 2 papers · 4 SOTA

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

Audio1 result

Speech RecognitiononCHiME-6 dev_gss12
Word Error Rate (WER)· 2020-04-22
55
best: 26 (ConformerXXL-PS + G-Augment)
SOTA
Towards a Competitive End-to-End Speech Recognition for CHiME-6 Dinner Party Transcription arXiv:2004.10799

Speech1 result

Audio-Visual Speech RecognitiononLRS3-TED
Word Error Rate (WER)· uses extra data· 2019-11-08
4.5
best: 0.74 (MMS-LLaMA)
SOTA
Recurrent Neural Network Transducer for Audio-Visual Speech Recognition arXiv:1911.04890

Computer Vision1 result

LipreadingonLRS3-TED
Word Error Rate (WER)· uses extra data· 2019-11-08
33.6
best: 12.8 (LP + Conformer)
SOTA
Recurrent Neural Network Transducer for Audio-Visual Speech Recognition arXiv:1911.04890

Natural Language Processing1 result

Natural Language TransductiononLRS3-TED
Word Error Rate (WER)· uses extra data· 2019-11-08
33.6
best: 12.8 (LP + Conformer)
SOTA
Recurrent Neural Network Transducer for Audio-Visual Speech Recognition arXiv:1911.04890