Google has added a number of languages to Google Translate including Zulu, Xhosa and Swahili and it reckons that it will translate these languages rather effectively.

The reason behind this claim is the firm’s own Google Neural Machine Translation (GNMT) system. This system allows Google’s neural network to translate sentences not only faster but more accurately.

In fact in a paper published last year Google said that its GNMT system reduced translation errors by an average of 60% when compared to the older method.

This incredible increase in accuracy is due Googles approached the translation technique. Principal scientist at Google, Blaise Aguera y Arcas, explains that to improve the neural network the AI was fed text that had already been translated across many languages.

This helped the machine learn how exactly it should execute translations so that we get the best result possible. In our testing we’ve found that the machine is alarmingly accurate, even if it does struggle a bit with context.

This functionality extends to the Google Translate app which will allow users to snap a photo with text and translate it right on their handset. The results are pretty good even when you’re looking at things like grammar and tense.

We snapped a picture, Google translated it and it’s pretty accurate.

You can test it out for yourself by downloading the app for Android or iOS.

We’re interested to hear your experience with the new improved Translate so why not give it a download and tell us what you think of Google’s new machine learning tricks.


Brendyn Lotz writes news, reviews, and opinion pieces for Hypertext. His interests include SMEs, innovation on the African continent, cybersecurity, blockchain, games, geek culture and YouTube.