Thousands of indigenous languages across the world are in danger of extinction, as elders who preserved them die out and dominant languages like English extend their grip. Some indigenous technologists believe that AI has a role to play in the preservation process. In New Zealand, the team at Te Hiku Media built their own automatic speech recognition (ASR) model for the indigenous Te Reo language, building on 30 years worth of archival footage and solicited audio clips from community members. Their speech-to-text model now transcribes Te Reo with 92% accuracy, outperforming similar attempts by major international tech companies.
[time-brightcove not-tgx=”true”]
Peter-Lucas Jones, the company’s CEO, who is from Kaitaia, New Zealand, is adamant that this technology is owned by and directly benefits his own Māori community, which has long been marginalized and oppressed. “In the digital world, data is like land,” he says. “If we do not have control, governance, and ongoing guardianship of our data as indigenous people, we will be landless in the digital world, too.”
This year, Te Hiku is working with Radio New Zealand to provide better Te Reo Māori transcripts; the New Zealand Qualifications Authority for language exams; and a Hawaiian translation organization, to see how the collaborations within the larger Polynesian language group might benefit everyone involved.
Jones hopes that Te Hiku’s efforts go far beyond sheer language acquisition. “If we can enhance the way that someone respects another language, we too may be able to shape the way they consider other people,” he says. “This is about Mana motuhake: being in control of one’s destiny, and growing an economic future for your people when that has largely been diminished through the process of colonization.”
Leave a comment
