Facebook has used the MIT Technology Review’s EmTech Digital conference in San Francisco to reveal its latest advances in the field of online translation. But what might those developments mean for the professional human translation industry?
Facebook’s online translation – the numbers
Facebook uses its translation tool to deliver a staggering 2 billion text translations per day. The tool covers 40 languages, in 1,800 directions. The result is that almost 50% of Facebook users see translated text each month – that’s 800 million users per month.
Until December 2015, Facebook was using Microsoft Bing to generate its translations. However, as an early adopter and pioneer of online translation the company was keen to implement its own system, which it has now done.
How does the Facebook online translation tool work?
Facebook’s online translation tool uses machine learning and artificial intelligence to deliver automated translation of users’ posts and comments. Interestingly, the system doesn’t just translate words, but also infers context from the text and images that it is translating.
That, according to Facebook, was why they moved away from Bing. Alan Packer, Facebook’s director of engineering for language technology, observed that Bing, “didn’t do well on slang, idioms, and metaphors. We really needed to train on our own data.”
Thus the Facebook online translation tool moved away from the more formal translation offered by sites like Bing and towards using an understanding of natural language, a process enhanced by the company’s acquisition of Wit.ai in 2015. Facebook’s Alan Packer comments:
“We did our own internal bake-off. When we could show it was better than Bing [for a specific language to language translation], we would turn Bing off and replace it with our own service.”
The future of online translation
Facebook is increasingly confident in both the value and the accuracy of its translation tool. The goal is to connect users around the world, irrespective of language barriers. That confidence will see the site move from a ‘see translation’ option to automatic translation backed by a ‘see original’ option.
The implications of the developments are huge. If machine translation can truly master the idioms, slang and abbreviations used in everyday language in order to produce accurate translations into any language, it won’t just be people posting videos of cute cats who benefit.
Companies will no doubt be queuing up to use the translation tool, which will enable them to effortlessly reach out to clients around the world. Rather than paying to have their website translated, companies can pay for a plug-in that will allow users to view the site in their native language, whatever that language may be. The company can add to the website, for example with blog posts and link it to their social media feeds, with the confidence that users will be able to connect with them across all platforms.
What does this mean for professional human translation?
Online natural language translation of this nature and on this scale is a big step forward for the machine translation sector. Google and Microsoft/Skype are racing to achieve further breakthroughs in order to deliver natural language translation to users around the globe.
So will the human translation industry be impacted? Quite possibly. Imagine a company that currently uses human translation for its business letters. If that business could instead upload its letters to a file sharing site that also offered translation, the documents could instantly be available in any language, as soon as they were uploaded. And if the same translation tool is applied to instant messaging – and in due course to email systems – then the translation could take place between one person clicking ‘send’ and the other receiving it in their inbox.
Key to it all is the quality of the translation and that is perhaps the most exciting part of Facebook’s revelation – instant, natural language translation could be a real game changer for the translation industry as we currently know it.
How long do you think it will take before machine translation is of good enough quality and widely available enough to impact the professional translation industry? Share your thoughts via the comments.