Just in case you need more proof that most machine translation tools don’t quite cut the mustard, the latest attempt by social media giant Facebook to incorporate machine translation (MT) into its platform fails miserably with most languages.
In an effort to help pages connect better with their fan base—often found scattered across the globe—Facebook recently introduced machine translation, powered by Bing. With just one click, users can get an automatic translation of status updates and comments. Facebook plans to roll out this feature to all profiles (not just pages) in the near future.
There’s just one problem: most of the translations are unintelligible. Posts on social media sites like Twitter and Facebook are rife with slang, and Bing’s machine translation tool simply isn’t up to task. An analysis of Bing’s performance by the site Lexiophiles shows that Bing only gets it right about 50% of the time, leaving users confused and, at worst, possibly misled by faulty translations. Interestingly, out of the ten languages tested, posts translated from Spanish to English fared worst of all, with less than 10% of the rendered text considered intelligible.
For greater accuracy, Facebook will also be implementing a feature that allows bilingual users to offer an alternative translation. If other users endorse the accuracy of the crowdsourced translation, it will take the place of Bing’s original translation the next time the “Translate” option is clicked. Page administrators will be able to manage crowdsourced translations through a “manage translations” link below the posts on pages they control.
One of the great arguments in favor of MT has always been that it can at least offer users a gist of the conversation when no other means of translation is available. It seems that, at least for now, Bing’s tool doesn’t even offer that to Facebook’s users.