The former Facebook language translation model runs a bit slow, so the network is revamping how it processes foreign languages to increase speed and accuracy…
Today, Facebook announced it is implementing a new form of language translation. And, it works at a quicker speed. Plus, it’s more accurate than its predecessor. What’s more, the new system helps the social network catch problems.
Facebook Introduces Faster, More Accurate Language Translation
Facebook now uses multilingual word embedding, which performs between 20 and 30 times faster than the natural language processing it used previously. Until this time, the company states the time it took to translate a new language ran about the time it would take to build a new website. Now, the social giant uses language vectors which group together words with the same meaning. (It works much like how humans process a word they are familiar with because they don’t have to think about it.)
The new Facebook language translation will not only help users, but the network, as well. Because of it, Facebook can release features more smoothly, since it no longer must wait long periods for translations. The new tech will likewise help the company detect content policy violations that might otherwise go unnoticed. Additionally, the new machine translation improved accuracy to 95 percent. That means the company can more swiftly spot intent and purpose behind non-English content.
Currently, Facebook is also working on idiom translation, along with other cultural nuances that might not translate between languages easily. Right now, the social site is using embedding with demonstrable success in German, French, and Spanish. Plus, it’s working to put forth the embedding system into the other one hundred languages on the platform.