Tired of having to piece together a coherent phrase after using Google Translate? Worry not, because Google Translate just got a lot better thanks to the company’s decision of employing powerful machine-learning techniques.
On Tuesday, the tech company announced that the quality of translation for nine languages has significantly improved after machine learning technique were added. So far, according to Google, the application will be able to translate more accurately words and phrases in Korean, Japanese, Chinese, Turkish, Spanish, Portuguese, German, French, Spanish, and English.
However, as the company pointed out, this only the beginning of a process that can potentially revolutionize the way we interact with translation engines. Barak Turovsky, the chief of Google Translate, declared that since the update rolled out at the beginning of this week, it is being used to process approximately 35 percent of all online translation requests.
Moreover, Turovky added that if the new translation technique proves its worth, it will be upgraded to include 103 languages. Turovsky also declared that in only a couple of day, the company managed to innovate more in terms of accurate translations than it did in the last decade.
It would seem that neural machines are becoming popular these days. While DeepMind wrestles with the toughest StarCraft II players and ‘Quick, draw!’ strives to figure out your doodles, Google Translate will now use its ‘imagination’ and the lesson it learned in order to give you an accurate translation.
Remember how hard it was to translate something from German or French into English? The words were correctly translated, but the whole sentence looked like it was written by someone who tried to drunk-text his ex-girlfriend.
Well, it would seem that thanks to Google’s latest update, you won’t need to worry about syntax. According to Google, the software’s accuracy was increased after the company implement several tensor processor units.
These processors will increase the speed of predictions up to three times. So, by the looks of it, it’s bye-bye jumbled texts and hello Oxford-like translations.
So far, the company will only test out the neural machine on this nine languages, but the head of Google Translate is confident that in a year’s time the app will employ the same technique for all 103 supported languages.
To see the improvements for yourself, fire up the web-based application or used the mobile application. Be sure to use one of the languages listed above.
Image source: Pixabay