Kaituhi by Te Hiku Media
A Māori-led speech recognition system transcribes te reo Māori and keeps language data under local control.
About
Te Hiku Media built Kaituhi for automatic transcription of te reo Māori, New Zealand English, and conversations that mix both languages. Iwi radio stations, researchers, and families can use it to create subtitles and searchable archives. The New Zealand government cites it as an example of AI for the preservation of an indigenous language. The creators do not treat language recordings as free-to-harvest data: users retain control, and new training data is only used with permission. Automatic transcription remains a process that people must verify, especially in cases of noise, dialects, and rare words.
AI factor
A proprietary speech model converts audio into text and is trained on controlled Māori language data. Māori staff and users improve the transcriptions and, according to the principle of kaitiakitanga, decide who may use language data and for what purpose.
Sources
Inspiration sourced from AI For Good. Not an Arnaldus customer case.