A research group led by Professor Tatsuya Kawahara of the Graduate School of Informatics, Kyoto University has developed artificial intelligence (AI) that automatically recognizes and synthesizes Ainu language, which is recognized as an extinction crisis language.This is the first time to build a voice recognition / synthesis system for Ainu language, and it seems to be effective for the transmission and learning of Ainu language.
According to Kyoto University, the research group received voice data of about 10 hours of folk tales from a total of 40 people provided by the Ainu Museum in Shiraoi Town, Hokkaido, and the Nibutani Ainu Culture Museum in Biratori Town, Hokkaido. AI achieved a phonetic recognition rate of 94% and a word recognition rate of 80% in the deep learning used as.
As a result, the work that takes one day by human hands can be almost completely automated for one hour of data, and the construction of the Ainu language archive has become much more efficient.
Furthermore, among the provided voice data, voices were synthesized by deep learning of AI for those who talked for 1 hours or more per person. When it was demonstrated at the Ainu Language Archive Study Group held in Shiraoi Town in September, it was well received by museum officials.
Ainu is the language of the Ainu who lived in Hokkaido, Sakhalin, and the Kuril Islands, and has been transmitted by oral tradition.Scientific research started after the Meiji era, but it is said that speakers have already disappeared in Sakhalin and the Kuril Islands.The number of speakers in Hokkaido decreased to 1996 in the 15 survey and 2007 in the 10 survey, and UNESCO confirmed in 2009 that it was in danger of a very serious disappearance.
