From NVIDIA:

Indigenous languages are quickly disappearing, with 3,000 at risk of disappearing by the end of the century. As part of the movement to save these languages, Te Hiku Media is using trustworthy AI to preserve and revitalize the Maori language, te reo. They have developed automatic speech recognition (ASR) models with 92% accuracy using NVIDIA technology.

Te Hiku Media, a broadcaster focused on the Maori people’s indigenous language, is using trustworthy AI to develop automatic speech recognition (ASR) models for the Polynesian language, te reo. These models, built with NVIDIA technology, transcribe te reo with 92% accuracy and can also transcribe bilingual speech with 82% accuracy.

Te Hiku Media began as a radio station to ensure te reo had space on the airwaves and has since evolved into a digital platform called Whare Kōrero. Now, around 20 Māori radio stations use and upload their content to Whare Kōrero, which holds over 30 years’ worth of content with about 1,000 hours of te reo native speakers.

Te Hiku Media has used trustworthy AI to transcribe te reo, Māori language, with a 92% accuracy and establish Kaituhi, a te reo Māori transcription service available online. It initiated a campaign to collect labeled speech samples, getting 2,500 people which led to 300 hours of labeled speech data, and contributed to training the ASR models.

Te Hiku Media is using trustworthy AI to transcribe te reo Maori language with a 92% accuracy rate, and its efforts have influenced similar ASR projects by Native Hawaiians and the Mohawk people in southeastern Canada. The organization’s use of NVIDIA technology has been game-changing and an inspiration for other indigenous groups.



Read more: Māori Speech AI Model Helps Preserve and Promote New Zealand Indigenous Language