List your company
ResearchFormerly WizWand

Automatic Speech Recognition on LRS2-BBC (test)

0.056WERWhisper (real label)
0.030080.205040.380.55496Sep 6, 2018Aug 19, 2019Aug 1, 2020Jul 15, 2021Jun 27, 2022Jun 10, 2023May 23, 2024
Updated 8mo ago

Evaluation results

MethodLinks
Whisper (real label)Adaptation strategy=Finetuning, Supervision type=Supervised2024.050.056
STARAdaptation strategy=STAR, Supervision type=Unsupervised2024.050.07
TOKreweight (A_t)Adaptation strategy=Self-training, Token-level re-weighting=A_t, Supervision type=Unsupervised2024.050.074
UTTfilterAdaptation strategy=Self-training, Filtering=Utterance-level, Supervision type=Unsupervised2024.050.076
TOKreweight (G_t)Adaptation strategy=Self-training, Token-level re-weighting=G_t, Supervision type=Unsupervised2024.050.079
TM-CTCModality=Audio-Visual, External Language Model=true2018.090.082
Whisper (self-train.)Adaptation strategy=Self-training, Supervision type=Unsupervised2024.050.083
TM-seq2seqModality=Audio-Visual, External Language Model=true2018.090.085
Whisper (frozen)Adaptation strategy=None, Supervision type=Zero-shot2024.050.085
TM-seq2seqModality=Audio-Visual, External Language Model=false2018.090.094
TM-seq2seqModality=Audio, External Language Model=true2018.090.097
TM-CTCModality=Audio, External Language Model=true2018.090.101
TM-seq2seqModality=Audio, External Language Model=false2018.090.105
TM-CTCModality=Audio-Visual, External Language Model=false2018.090.137
TM-CTCModality=Audio, External Language Model=false2018.090.153
Google S2T+Modality=Audio, External Language Model=false2018.090.209
TM-seq2seqModality=Video, External Language Model=true2018.090.483
TM-seq2seqModality=Video, External Language Model=false2018.090.498
TM-CTCModality=Video, External Language Model=true2018.090.547
TM-CTCModality=Video, External Language Model=false2018.090.65
WASModality=Video, External Language Model=false2018.090.704