The Microsoft 2016 Conversational Speech Recognition System

12 Sep 2016W. XiongJ. DroppoX. HuangF. SeideM. SeltzerA. StolckeD. YuG. Zweig

We describe Microsoft's conversational speech recognition system, in which we combine recent developments in neural-network-based acoustic and language modeling to advance the state of the art on the Switchboard recognition task. Inspired by machine learning ensemble techniques, the system uses a range of convolutional and recurrent neural networks... (read more)

PDF Abstract

Code


No code implementations yet. Submit your code now
TASK DATASET MODEL METRIC NAME METRIC VALUE GLOBAL RANK RESULT LEADERBOARD
Speech Recognition swb_hub_500 WER fullSWBCH VGG/Resnet/LACE/BiLSTM acoustic model trained on SWB+Fisher+CH, N-gram + RNNLM language model trained on Switchboard+Fisher+Gigaword+Broadcast Percentage error 11.9 # 2
Speech Recognition Switchboard + Hub500 VGG/Resnet/LACE/BiLSTM acoustic model trained on SWB+Fisher+CH, N-gram + RNNLM language model trained on Switchboard+Fisher+Gigaword+Broadcast Percentage error 6.3 # 4
Speech Recognition Switchboard + Hub500 RNNLM Percentage error 6.9 # 6
Speech Recognition Switchboard + Hub500 Microsoft 2016 Percentage error 6.2 # 3