advertisement
advertisement

This new app from Google is designed to preserve the words of fading languages

The 10 initial languages featured on the app include Creole, Maori, Sicilian, and Yiddish.

advertisement
advertisement
advertisement

For many reasons, including globalization and cultural assimilation, a handful of languages, such as English, Spanish, and Mandarin, are dominating the world’s linguistic landscape—and that often comes at the expense of older and less popular dialects, which slowly fade out. It’s estimated that a language goes extinct every 14 days; almost half of the world’s 6,000 to 7,000 languages are endangered. UNESCO has a scale for threatened languages, called the Atlas of the World’s Languages in Danger, where tongues range from vulnerable to critically endangered (the rung on the scale right before “extinct”).

advertisement
advertisement

This modern-day reality creates a distressing sense of loss for many people who understandably want to preserve their cultural heritage and keep their family traditions from fading into obsolescence. That’s why Google Arts & Culture is deploying its machine-learning tech to allow anyone in the world to easily find words for common objects in 10 of these endangered languages. Through image detection technology, and partnerships with language preservation groups around the world, the project is curating an ever-expanding glossary of words, to be a source of hope for those with attachments to a historic culture, or of fun for those who simply want to learn about a new language.

[Image: Google Arts & Culture]
The free app is part of Google Arts & Culture’s mission to “democratize access to the world’s arts and culture,” says Chance Coughenour, the Google division’s head of preservation, which it does with the help of 2,500 partners in 80 countries. The division first started by digitizing pieces of museum art for public online access, and it’s now branched into using its tech to help preserve “intangible heritage,” or “the ephemeral part of heritage that is at risk of being lost or endangered,” Coughenour says.

Users can pull up the app, called Woolaroo, on their mobile browsers and take a photo of any object, or a scene containing several objects. Google’s Cloud Vision API, its image recognition system that’s used for such programs as Google Lens, analyzes the photo based on its machine learning data from having processed millions of images, explains Ian Pattison, head of retail engineering at Google Cloud U.K. The app will generate suggestions for each object in a photo—along with the translation for that word in the chosen language, plus an audio pronunciation of that word.

advertisement

[Image: Google Arts & Culture]
The 10 languages include two Italian languages: Sicilian and Calabrian Greek, a dialect of Greek still spoken in some villages in the southern region of Calabria (the toe of the Italian boot) by about 2,000 people. There’s Louisiana Creole, a French-based language, spoken by about 7,000 in certain Louisiana parishes. There’s Nawat (or Pipil), a language found in El Salvador spoken by 200 people, labeled by UNESCO as critically endangered, the most threatened level before extinction. In order to be accessible to a wide range of people, the app works in English, Arabic, Spanish, French, and Italian. Someone in North Africa looking to learn new words in the Berber language of Tamazight, for instance, would likely already know either Arabic or French.

[Image: Google Arts & Culture]
Google is partnering with groups in the 10 native countries, each of which has an interest in preserving the language. For a dialect found in the Guangxi province of China, Yang Zhuang, the partner is the Museum of Ethnic Cultures at the Minzu University of China; for Yiddish, it’s the National Yiddish Theater in New York; and for the Polynesian language of Rapa Nui, Google partnered with a council of elders on Easter Island. These partners helped curate the word lists and provide the translations and pronunciations, and they’ll also help conduct improvements. Users can request to correct pronunciations or to add more words if the objects they photograph don’t come back with suggestions.

Woolaroo is an evolving project, and Coughenour says they’re working on adding more languages to the mix. The aim isn’t to fully learn a new language, in the mold of apps such as Duolingo, since the app is only feeding users simple nouns. “You’re certainly not going to learn how to speak the language,” he says. But it’s a fun way to learn new words, and for people to connect with cultures. It could even be used as a classroom aid, for instance for Kiwi schoolchildren who learn Maori as part of their curriculum.

advertisement

Incidentally, Woolaroo is a word in Yugambeh, an Australian Aboriginal language now spoken in Queensland by only about 100 people, meaning “shadow”—the closest word in the language to “photo,” Coughenour says. Ironically, for a project leveraging tech to bolster dying dialects, “Many indigenous languages around the world don’t necessarily have words that are direct translations to new technology that exists today.”