Improvement of naturalness of learners' spoken Japanese by practicing with the Web-based prosodic reading tutor, Suzuki-kun

Author(s): Nobuaki Minematsu, Hiroko Hirano, Noriko Nakamura and Koji Oikawa


It is known that learning prosodic control for speaking Japanese is effective to reduce syntactic and lexical ambiguity of learners' spoken Japanese and to improve its naturalness and comprehensibility. In conventional curriculum, however, prosody training has not been provided satisfactorily for learners partly because teaching materials for prosody training are limited. In our previous studies, we built a web-based system that, for any given text, can illustrate the prosodic control required to read that text in Tokyo Japanese and also provide a high-quality synthetic voice for that text based on the visualized prosody. Although this system is currently used by many learners, assessment of the system was done only in terms of users' satisfaction. In this study, the effectiveness is examined quantitatively by using eighty Chinese learners of Japanese. We compare the improvements of naturalness realized by practicing 1) with synthetic voices, 2) with visualized prosody, and 3) with both of them. Experimental results show that simultaneous use of auditory and visual instructions is the most effective and that visual prosody can help learners much more than auditory prosody.