In what is another interesting example of just how big Facebook has become over the years, the firm now handles all of their own translations themselves, using a nifty little AI and their own machine learning techniques. Previously, Facebook had partnered with Bing’s translation service to help international users read posts from all around the world in their native tongue without moving away from Facebook. Over the years however, Facebook began to work on their own solution and since December of 2015, Facebook started to go it alone, and all of their translations since then have been managed by their own software.
Alan Packer, Facebook’s Director of Engineering for language technology was recently speaking at the Emtech Digital conference, held by MIT, and he had some interesting figures to share. One of which is that Facebook’s AI responsible for handling translation can understand over 40 different languages and dialects and that 800 Million Facebook users all over the world benefit from the new system, with Facebook translating 200 Million lines of text every single day. The reason the team moved away from using Bing’s translation software is that Microsoft’s option was designed to work on proper language, not necessarily the way that humans talk to one another, and struggled with slang, for instance. That’s when Facebook got to work creating an AI that could automatically translate text in a similar fashion to a human translator, just one that worked much, much quicker. This has led to Facebook often automatically translating text if the AI’s response meets a certain threshold, with a “See Original” button being used instead of a translate button.
Translation appears to be one area that Google, Microsoft and now Facebook are all competing in, as more and more nations come online with faster speeds, users are demanding access to a more varied amount of content, and it appears as though translation is one of the few key features a tech firm needs to get right in order to appeal to a wide audience these days.
It’s interesting to see how Facebook approaches artificial intelligence, and with 1.6 Billion users on Facebook, they arguably have more human interaction to work with while training these AIs than practically anyone else. This translation system is just one example that has helped Facebook improve their network on a global scale, and it’s something that Facebook hopes will continue to break down barriers and naturally increase people’s interaction on Facebook no matter where in the world they are.