Embedded Files
Measurement of listening disfluency (LD) and
If listening disfluency can be measured easily, it will change many aspects of language learning,
because learners can know how fluently they are understood by others .
Minematsu-Saito lab., Graduate School of Engineering, UTokyo, Japan
Project leader : Nobuaki Minematsu, Professor, The University of Tokyo
Aim of the project
Aim of the project
Listening is a mental process, and measurement of listening disfluency (LD) may require expensive techniques of brain sensing. Are there any good alternatives to measure LD objectively with a reasonable cost? In this project, LD is converted to acoustic signals, and they are captured with a microphone as sequential data temporally aligned with the presented audio to the listener.
Any language learner wants to make him/herself understood easily by various others. If s/he can estimate the listener's mental state while talking, s/he may be able to control his/her speaking manner to become more intelligible to the listener.
Shadowing-based acoustic measurement of LD
Shadowing and mirroring
To assess learners' listening skills
To reduce learners' listening disfluency
To assess learners' speaking skills
To score the intelligibility for learners' speech
Prediction of learners' LD using machine learning
Sequence-to-sequence voice conversion is applied to predict raters' LD.
Send emails to shadowing [ATMARK] gavo.t.u-tokyo.ac.jp
Minematsu-Saito lab. of Graduate School of Engineering, UTokyo, Japan
Page updated
Report abuse