Google Woolaroo
"Woolaroo was co-developed by RUSH, one of New Zealand’s most innovative digital product studios. RUSH took on the project to help use innovative technologies like Machine Learning, Machine Vision and the power of the Cloud in a project that aligned so closely to its purpose statement."
The project
UNESCO predicts the loss of as many as 3,000 indigenous languages by the end of this century. RUSH worked on the successful use of image recognition to translate photos into te reo Māori in 2019 - and now with Google, sought to expand this technology to the thousands of other indigenous languages around the world that need preserving.
Expertise
Accessibility
Backend Development
Cloud / Systems Integration
Frontend Development
Machine Learning
UI Design
476M
People globally using indigenous languages
3000
Indigenous languages predicted to be lost by the end of the century
17
Endangered languages supported on Woolaroo
Build, measure, learn
In 2019 we worked with Spark and Colenso BBDO on the successful use of image recognition to translate photos into te reo Māori with the app Kūpu. So how might we expand this technology to the thousands of other indigenous languages around the world that need preserving?
Google Arts & Culture wanted to broaden the experiment to build an open source digital product designed to scale across the globe to save all languages on the brink of extinction.
In a closed mobile application setup, proprietary data access must be granted from each contributing community - easy when there’s only one language source. But we want to add dozens, eventually hundreds, or thousands. Woolaroo was built on an open-source platform that ensures language resources remain the property of the source provider through licensed APIs, and indigenous communities retain data sovereignty.
"Since Woolaroo launched, our partners have been using the app as a resource for teaching next generations of speakers in their communities."
Preserving languages with machine learning
Woolaroo was built on an open-source platform that ensures language resources remain the property of the source provider through licensed APIs, and indigenous communities retain data sovereignty. To use Woolaroo, users select their primary language, then scroll for an Indigenous language to learn. Woolaroo uses advanced machine learning and Cloud Vision APIs to understand objects in an image and immediately translate them into the target language. If a user takes a picture of a tree, Woolaroo will display the Yugambeh word “tullei,” as well as alternatives if multiple objects are detected.
Language is ancient tech
The real magic is that Woolaroo is constantly learning and improving. Indigenous speakers can suggest corrections, add new words and audio recordings for understanding of pronunciation. This creates a collaborative and growing word list for each language and keeps the spoken language alive. Users also have the chance to learn more about the background and culture of each language. A July 2022 regional launch was done in partnership with the UNESCO Decade of Indigenous Languages Secretariat, and continued work with them will highlight Woolaroo amongst their network of institutions.
See it in action
This web-based platform brings a simple UI to mix primary languages and Indigenous languages. Designed for extreme simplicity and accessibility, the UI scales on global devices, and cloud-based fonts easily crossover between all languages, including left-to-right languages.
This web-based platform brings a simple UI to mix primary languages and Indigenous languages. Designed for extreme simplicity and accessibility, the UI scales on global devices, and cloud-based fonts easily crossover between all languages, including left-to-right languages.
Ready to make better happen?
