Google develops app to learn dying languages ​​with image recognition

Google has developed a new app through the Arts & Culture project to get users excited about endangered languages. The app uses machine learning to recognize objects in photos, giving the word for an object in the dying language.

With the Woolaroo app, users can now learn words from ten endangered languages ​​by taking a picture of an object, such as a tree. The app uses image recognition and machine learning to recognize the object and display the corresponding word in Louisiana Creole, Calabrian Greek, Māori, Nawat, Tamazight, Sicilian, Yang Zhuang, Rapa Nui, Yiddish, and Yugambeh. Very few speakers of these languages ​​remain.

The app has been developed as open source and language communities can also save a language in the app, add words and audio recordings of those words, edit and delete words and give context to those languages. Most of the languages ​​are barely spoken. With the app, Google hopes to make the languages ​​easily accessible to new generations and to preserve them.

Woolaroo uses the Google Cloud Vision AI API. That api can recognize images using machine learning, in order to give the correct word in real time to an object that the app recognizes. The api uses AutoML and pre-trained models to quickly categorize images.