SPE-50.2
HAVE BEST OF BOTH WORLDS: TWO-PASS HYBRID AND E2E CASCADING FRAMEWORK FOR SPEECH RECOGNITION
Guoli Ye, Vadim Mazalov, Jinyu Li, Yifan Gong, Microsoft, United States of America
Session:
Speech Recognition: Acoustic Modeling II
Track:
Speech and Language Processing
Location:
Gather Area C
Presentation Time:
Wed, 11 May, 20:00 - 20:45 China Time (UTC +8)
Wed, 11 May, 12:00 - 12:45 UTC
Wed, 11 May, 12:00 - 12:45 UTC
Session Chair:
Bo Li, Google
Session SPE-50
SPE-50.1: NON-AUTOREGRESSIVE ASR WITH SELF-CONDITIONED FOLDED ENCODERS
Tatsuya Komatsu, LINE Corporation, Japan
SPE-50.2: HAVE BEST OF BOTH WORLDS: TWO-PASS HYBRID AND E2E CASCADING FRAMEWORK FOR SPEECH RECOGNITION
Guoli Ye, Vadim Mazalov, Jinyu Li, Yifan Gong, Microsoft, United States of America
SPE-50.3: CONFORMER-BASED HYBRID ASR SYSTEM FOR SWITCHBOARD DATASET
Mohammad Zeineldeen, Christoph Lüscher, Wilfried Michel, Ralf Schlüter, Hermann Ney, RWTH Aachen University / AppTek, Germany; Jingjing Xu, Alexander Gerstenberger, RWTH Aachen University, Germany
SPE-50.4: Improving Factored Hybrid HMM Acoustic Modeling without State Tying
Tina Raissi, Ralf Schlüter, RWTH Aachen University, Germany; Eugen Beck, Apptek GmbH, Germany; Hermann Ney, RWTH Aachen Univeristy, Germany
SPE-50.5: AUDITORY-BASED DATA AUGMENTATION FOR END-TO-END AUTOMATIC SPEECH RECOGNITION
Zehai Tu, Jack Deadman, Ning Ma, Jon Barker, University of Sheffield, United Kingdom of Great Britain and Northern Ireland
SPE-50.6: DELIBERATION OF STREAMING RNN-TRANSDUCER BY NON-AUTOREGRESSIVE DECODING
Weiran Wang, Ke Hu, Tara Sainath, Google, United States of America