Kentaro Onda, first-year student at the Department of Electrical Engineering and Information Systems received IPSJ-SLP Best Presentation Award.

Name of award and short explanation about the award
IPSJ-SLP Best Presentation Award

About awarded research(activity)

Synthesis of Foreign Accented Speech by Self Supervised Learning Only Using Native Speech Corpora. Kentaro Onda, Joonyong Park, Nobuaki Minematsu, Daisuke Saito

Your impression & future plan

In today's world, everyone speaks foreign languages, yet both humans and speech recognition AIs still struggle to understand accented speech. One reason for this is the significant lack of training data for accented speech. However, with this method, it is possible to generate various combinations of "A accented B" only from native speech corpora. The synthesized speech samples can be used as listening materials or speech corpora. Furthermore, the technology used in this method is being considered to be used for learners’ pronunciation evaluation and difficulty estimation of educational materials. This research has just begun. Under the guidance of Prof. Minematsu and Prof. Saito, we will continue our research on accented speech with the key phrase "using only native speech corpora".