It is said that there are more than 7,000 languages in the world, yet only about one hundred are currently covered by AI. Many of the remaining languages are at risk of extinction. Since language serves as a vehicle for transmitting the culture of a people or region, the disappearance of a language can lead to the disappearance of a culture. In response to this, various initiatives have been launched to preserve, pass on, and revitalize endangered languages.
This presentation introduces the research on speech recognition and synthesis for the Ainu language that is being undertaken as part of these efforts to preserve endangered languages. With just 5–10 hours of speech data from a small number of speakers, we achieved over 90% recognition accuracy, contributing to the archiving of many audio sources. We also developed a speech synthesis system capable of generating recordings of traditional narratives for which no audio materials exist.
Looking ahead, we hope to extend this work to the Ryukyuan languages and various other dialects.