Mozilla has launched a project to create a machine translation system that operates on the user's browser. The project will integrate a standalone page translation engine into Firefox that does not rely on external cloud services and processes data solely on the user's system. The main goal of the development is to ensure user privacy and protect data from potential leaks when translating content on web pages opened in the browser.
The Bergamot project is being developed at Mozilla's Berlin office with the participation of researchers from universities in the United Kingdom, Estonia, and the Czech Republic. The development is funded by the European Union under a grant obtained through the program. The grant amount is three million euros. The project is set to last for three years. Mozilla for a machine learning systems specialist to participate in the development of an engine for translating from one language to another.
Related developments concerning the Bergamot project :
- The machine translation framework developed at the University of Edinburgh , built on a recurrent neural network. The framework is written in C++, can utilize GPU for accelerated learning and translation, and is licensed under the MIT License.
- The toolkit created at Charles University for processing information in natural language using sequential machine learning methods. The project applies the TensorFlow framework and can be used for rapid prototyping of machine translation and natural language classification systems. The code is available under the BSD license.
- The project prepared at the University of Sheffield , used for assessing and predicting the quality of machine translation systems.
- Developed at Mozilla () and ()
- The European Union-funded project , which accumulates a base of synchronized translations of various phrases in different languages that can be used for training machine learning systems. The basis of the project is the bot , which indexes multilingual websites and automatically finds identical texts presented in multiple languages. The database of parallel translation examples is being formed for 24 languages, including Russian ( for the Russian language occupies 637MB when compressed and includes over 12 million translation examples).
The development of Bergamot started in January of this year, but the project's work is still for public testing. Nevertheless, a few days ago the developers released a video demonstrating the initial prototype.
Let’s remember that Firefox already has a built-in mechanism for translating pages, but it relies on using external cloud services (supports Google, Yandex, and Bing) and is not enabled by default (to enable it in about:config, one should change the settings of "browser.translation"). The translation mechanism also supports automatic language detection when opening a page in an unknown language and displays a special indicator suggesting to translate the page. The prototype of the translation system developed within the Bergamot project uses the same interface for user interaction, but instead of calling external services, it launches a built-in handler.
Source: opennet.ru
