Facebook has released NLLB project (No language left behind). The main feature of this development is the coverage of over two hundred languages, including rare languages of African and Australian nations. In addition, Facebook used a new approach to the machine learning model, in which the translation takes place directly from one language to another, without intermediate translation into English.
This project will include all modern languages to provide means of communication for all people on Earth.
The model is distracted free of charge (for non -commercial use). To facilitate the designs of other companies, the Open Source code was used to test the assessment and quality of models, as well as to train models. The final model is offered in two versions: full and reduced.
Especially, NLLB It is a common model for all 200 languages and does not require any additional models for each language separately. The system can automatically recognize the original language and translate into a user language.
According to the project team, the system is almost one and a half times better than previous translation systems and better explains rare languages by seventy percent.
You can assess the quality of translation and read stories in rare languages here https://nllb.metademolab.com/
