The Mozilla Foundation has added an official translation tool to Firefox that does not rely on cloud processing to do its work, and instead runs the machine learning-based process on your computer.
This represents a huge step forward for a popular service strongly associated with giants such as Google and Microsoft.
A translation tool called Firefox Translations can be added to your browser. And you need to download some resources the first time you translate a language. You may also download enhanced templates if needed. But the actual translation work is done by your computer, not in a data center.
This is important because many people need translation in the browser while they are not connected to the Internet. This is also important because it aims to reduce reliance on cloud providers for a task that no longer requires their resources.
The translation feature is the result of the Bergamot project funded by the European Union. This project saw Mozilla collaborate with the University of Edinburgh, Charles University, University of Sheffield and University of Tartu to develop a suite of machine learning tools that would make offline translation possible.
This type of work is usually performed by GPU clusters in data centers, where large-format language models for user query translation are deployed.
But while the cloud-based tools from Google and Microsoft are accurate and fast with nearly unlimited computing power, there are primary privacy and security risks involved in sending your data to a third party for analysis and return.
Mozilla brings a translation tool to Firefox
For some, these risks are acceptable. While others prefer not to engage the online advertising giants if they don’t have to.
Offline translation makes sense for anyone concerned about the privacy implications of using a cloud translation provider.
The main limitation may be the lack of languages. Google Translate supports more than 100 languages. While Firefox translation is limited to dozens of languages, there is no Arabic language among them.
But it should be remembered that this is the first release of a project by a non-profit organization and a group of academics, and not a high-profile product from a multibillion-dollar Internet empire stretching across the globe.
Mozilla is actively seeking help by offering a training path to allow those interested to train new models. It also requests feedback to improve existing models.
