Language Science Talks: Dr. Nobuaki Minematsu, 'Speech Structure and Its Derivatives - What is the skeleton of spoken language? '
April 13, 2018, 11:30 am to 1:00 pm
Speech Structure and Its Derivatives - What is the skeleton of spoken language?
Prof. Nobuaki MINEMATSU
Graduate School of Engineering, the University of Tokyo
Abstract
Speech has huge acoustic variability in nature but humans use this fragile media as default media for language communication. With the help of DNN-based acoustic modeling, robustness of automatic speech recognition systems has been improved but it is still not comparable to robustness that humans exhibit.
How do human infants acquire super-robust or super-flexible capabilities of dealing with speech acoustics? The lecturer introduces a linguistically classical but technically new idea, speech structure, to this question. He also explains that, by assuming structural speech processing in humans, some mysterious behaviors of language disorder can be logically and easily predicted.
About Dr. Nobuaki Minematsu
Nobuaki Minematsu earned the doctor of Engineering in 1995 from the University of Tokyo. Currently, he is a full professor there. He has a wide interest in speech communication and covers from speech sciences to speech technologies. He also has expert knowledge on how to apply speech technologies for second language learning. He is a vice-chair of IEICE and a chair of Spoken Language Processing SIG of IPSJ, and is a member of IEEE, ISCA, SLaTE, IPA, APSIPA, IEICE, IPSJ, ASJ, etc.