Technical Program

Click on the icon to view the manuscript on IEEE XPlore in the IEEE ICASSP 2020 Open Preview.

SPE-L3: End-to-end Speech Recognition II: New Models

Session Type: Lecture
Time: Tuesday, 5 May, 16:30 - 18:30
Location: On-Demand
Virtual Session: View on Virtual Platform
Session Chair: Tara Sainath, Google
 
 SPE-L3.1: JOINT PHONEME-GRAPHEME MODEL FOR END-TO-END SPEECH RECOGNITION
         Yotaro Kubo; Google
         Michiel Bacchiani; Google
 
 SPE-L3.2: QUARTZNET: DEEP AUTOMATIC SPEECH RECOGNITION WITH 1D TIME-CHANNEL SEPARABLE CONVOLUTIONS
         Samuel Kriman; University of Illinois at Urbana-Champaign
         Stanislav Beliaev; University of Saint Petersburg
         Boris Ginsburg; NVIDIA
         Jocelyn Huang; NVIDIA
         Oleksii Kuchaiev; NVIDIA
         Vitaly Lavrukhin; NVIDIA
         Ryan Leary; NVIDIA
         Jason Li; NVIDIA
         Yang Zhang; NVIDIA
 
 SPE-L3.3: END-TO-END MULTI-TALKER OVERLAPPING SPEECH RECOGNITION
         Anshuman Tripathi; Google
         Han Lu; Google
         Hasim Sak; Google
 
 SPE-L3.4: END-TO-END MULTI-SPEAKER SPEECH RECOGNITION WITH TRANSFORMER
         Xuankai Chang; Johns Hopkins University
         Wangyou Zhang; Shanghai Jiao Tong University
         Yanmin Qian; Shanghai Jiao Tong University
         Jonathan Le Roux; Mitsubishi Electric Research Laboratories (MERL)
         Shinji Watanabe; Johns Hopkins University
 
 SPE-L3.5: HYBRID AUTOREGRESSIVE TRANSDUCER (HAT)
         Ehsan Variani; Google
         David Rybach; Google
         Cyril Allauzen; Google
         Michael Riley; Google
 
 SPE-L3.6: LIGHTWEIGHT AND EFFICIENT END-TO-END SPEECH RECOGNITION USING LOW-RANK TRANSFORMER
         Genta Indra Winata; Hong Kong University of Science and Technology
         Samuel Cahyawijaya; Hong Kong University of Science and Technology
         Zhaojiang Lin; Hong Kong University of Science and Technology
         Zihan Liu; Hong Kong University of Science and Technology
         Pascale Fung; Hong Kong University of Science and Technology