Second Language Studies: Acquisition, Learning, Education and Technology

Tokyo, Japan
September 22-24, 2010

Pronunciation Assessment System Using Structural Features

Masayuki Suzuki, Nobuaki Minematsu

The University of Tokyo, Japan

Pronunciation assessment systems have to be able to ignore speaker individualities like age or gender ( the size and length of the vocal tube). Our system ignore the speaker individualities by using approximately speakerinvariant features called structural features. The demo system can
1. record or log a history of vowel pronunciation training of each learner.
2. show which vowel to correct first to become like a teacher.
3. classify all the registered learners only with regard to pronunciation proficiency by ignoring gender, age, etc. very effectively. After the session, we may show a classification result of all the participants in our demonstration.

Full Paper

Bibliographic reference.  Suzuki, Masayuki / Minematsu, Nobuaki (2010): "Pronunciation assessment system using structural features", In L2WS-2010, paper D1-6.