List your company

Speech Recognition on BABEL (Assamese, Tagalog, Swahili, Lao)

27.2Assamese CERDinoSR
26.63630.44334.2538.057May 17, 2023

Evaluation results

MethodLinks
DinoSRPre-training steps=200k, Batch size (minutes)=162023.0527.219.414.622.8
XLSR-10Pre-training steps=250k, Batch size (minutes)=962023.0529.421.916.623.3
Seq-to-seq ASR without pre-trainingPre-training=without2023.0541.337.929.138.7