List your company
ResearchFormerly WizWand

Speech Translation on libri-trans (test)

23.1Detokenized BLEU (case-sensitive)End-to-end (w2v2-mBART)
10.93214.09117.2520.409Dec 18, 2020Apr 8, 2021Jul 29, 2021Nov 17, 2021Mar 9, 2022Jun 28, 2022Oct 18, 2022
Updated 8mo ago

Evaluation results

MethodLinks
End-to-end (w2v2-mBART)Learning Setup=Supervised learning + pre-training, Architecture=End-to-end2022.1023.1—
ST transf-base + AFSSystem Type=End-to-End, Architecture=Transformer Base, embedding size=512, Technique=Adaptive Feature Selection (AFS)2020.1217.218.6
NeurST ST transf-sCategory=End-to-End, Data setting=constrained2020.1217.218.7
TCEN-LSTMCategory=End-to-End, Data setting=constrained2020.1217.1—
ESPnet-ST ASR transf-s + CTC → MTCategory=Cascade, Data setting=constrained2020.1217—
NeurST ASR transf-s → MTCategory=Cascade, Data setting=constrained2020.1216.818.2
ESPnet-ST ST transf-sCategory=End-to-End, Data setting=constrained2020.1216.7—
NeurST ST transf-sSystem Type=End-to-End, Model Components=ST Transformer-Small2020.1216.317.8
ST BILSTMCategory=End-to-End, Data setting=constrained2020.1216.217
NeurST ASR transf-s → MTSystem Type=Cascade, Model Components=ASR Transformer-Small → MT2020.121617.4
Cascaded (ASR→TDN→MT)Learning Setup=Unsupervised learning, Architecture=Cascaded2022.1015.4—
End-to-end (w2v2-mBART)Learning Setup=Unsupervised learning, Architecture=End-to-end2022.1015.3—
Chung et al. (2019)*Learning Setup=Unsupervised learning, Architecture=not specified2022.1012.2—
End-to-end (S2T Transformer)Learning Setup=Supervised learning, Architecture=End-to-end2022.1011.4—
LUTCategory=End-to-End, Data setting=constrained2020.12—17.8
ST transf-s (Liu et al., 2019)Category=End-to-End, Data setting=constrained2020.12—14.3
ST transf-s (Wang et al., 2020c)Category=End-to-End, Data setting=constrained2020.12—16
ST transf-s + curriculum pre-trainingCategory=End-to-End, Data setting=constrained2020.12—17.7
ST transf-s + KDCategory=End-to-End, Data setting=constrained2020.12—17