Technical Program

MLSP-L3: Audio, Speech, and Natural Language Processing

Session Type: Lecture
Time: Wednesday, May 15, 08:30 - 10:30
Location: Auditorium 2
Session Chair: Zheng-Hua Tan, Aalborg University
 
MLSP-L3.1: HMM-BASED APPROACHES TO MODEL MULTICHANNEL INFORMATION IN SIGN LANGUAGE INSPIRED FROM ARTICULATORY FEATURES-BASED SPEECH PROCESSING
Manuscript Link:  Click here to view manuscript on IEEE Xplore
         Sandrine Tornay; Idiap Research Institute
         Marzieh Razavi; Telepathy Labs GmbH
         Necati Cihan Camgoz; University of Surrey
         Richard Bowden; University of Surrey
         Mathew Magimai.-Doss; Idiap Research Institute
 
MLSP-L3.2: SPEECH EMOTION RECOGNITION USING MULTI-HOP ATTENTION MECHANISM
Manuscript Link:  Click here to view manuscript on IEEE Xplore
         Seunghyun Yoon; Seoul National University
         Seokhyun Byun; Seoul National University
         Subhadeep Dey; Idiap Research Institute
         Kyomin Jung; Seoul National University
 
MLSP-L3.3: HOW TRANSFERABLE ARE FEATURES IN CONVOLUTIONAL NEURAL NETWORK ACOUSTIC MODELS ACROSS LANGUAGES?
Manuscript Link:  Click here to view manuscript on IEEE Xplore
         Jessica Thompson; Université de Montréal
         Marc Schönwiesner; University of Leipzig
         Yoshua Bengio; Quebec Institute for Learning Algorithms (Mila)
         Daniel Willett; Nuance Communications
 
MLSP-L3.4: GRADIENT-BASED ACTIVE LEARNING QUERY STRATEGY FOR END-TO-END SPEECH RECOGNITION
Manuscript Link:  Click here to view manuscript on IEEE Xplore
         Yang Yuan; Yonsei University
         Soo-Whan Chung; Yonsei University
         Hong-Goo Kang; Yonsei University
 
MLSP-L3.5: SEMI-SUPERVISED TRAINING FOR END-TO-END MODELS VIA WEAK DISTILLATION
Manuscript Link:  Click here to view manuscript on IEEE Xplore
         Bo Li; Google, Inc.
         Tara N. Sainath; Google, Inc.
         Ruoming Pang; Google, Inc.
         Zelin Wu; Google, Inc.
 
MLSP-L3.6: STOCHASTIC ADAPTIVE NEURAL ARCHITECTURE SEARCH FOR KEYWORD SPOTTING
Manuscript Link:  Click here to view manuscript on IEEE Xplore
         Tom Véniat; Sorbonne Université
         Olivier Schwander; Sorbonne Université
         Ludovic Denoyer; Facebook