Abstract of the lecture:
With recent advancement of speech technology, learners' speaking behaviors are analyzed, and corrective feedback is provided. Here, pronunciation errors, dysfluency, speaking breakdown, etc are automatically detected and used for corrective feedback. Is it also possible to detect listening breakdown using speech technology? Listening itself does not generate any acoustic signals and thus cannot be observed acoustically. In this talk, however, we will show that listening breakdown can be detected acoustically with speech technology by asking listeners to conduct a speech task while listening, called shadowing. Listening breakdown of evaluators, who are listening to L2 speech, is used to quantify the degree of intelligibility of that L2 speech objectively. Listening breakdown of learners, who are listening to L1 speech, is used to clarify their weakness of listening. Further, we show that our method of observing listening breakdown of learners can effectively enhance their listening skills.
Name of the speakers:
Prof. Nobuaki Minematsu (Graduate School of Engineering, The University of Tokyo)
Prof. Noriko Nakanishi (Faculty of Global Communication, Kobe Gakuin University)