Cookies on KBL website

To improve our website, we use Google Analytics cookies. These small pieces of data placed in your browser show us some of your activities on our website (such as which pages you’ve visited, etc.) and allow us to measure audience on the website. For more information, please visit our Website Data Protection Policy


25 April 2017

Lost in Translation 

Star Trek’s “universal translator” is no longer science fiction.

Digital translation tools have advanced significantly since Google introduced its instant translation engine in 2006.

The product turning the most heads is not even available yet. The app-based Pilot earpiece – which its creators are dubbing “the world’s first translating earpiece” – is meant to provide real-time translation straight into your ear. Some, $4.5 million worth of pre-orders have already been placed for these $249 earbuds.

Meanwhile, iTranslate is a phone app that provides real-time verbal translation – with over 60 million downloads, it is clear that people all over the world are particularly interested in instant translation.

Considering the steady pace of globalization, the $29 billion language technology market is poised to swell as demand surges. The industry is about to be disrupted by progress in the development of a rapidly evolving technology called Neural Machine Translation (NMT).

NMT feeds large sets of data through layers of interconnected processors modeled after the brain’s network of neurons. It is a significant step up from the phrase-based systems that most automated translators currently use; instead of dividing sentences into phrases and then individually translating them, NMT can analyze the wider context of the content.

This system even trains itself to improve, and reduces translation errors by up to 87% compared to current web-based tools.

Although NMT is a game-changer for automated tools, and integral to products like the Pilot, it is unlikely to have a major impact on the broader business of translation.

In the end, while developments in these tools are great for general translations, they are still far from replacing human translators. Although the technology can now take in the wider context, it is still far from being capable of understanding and aptly translating all the nuances of our distinct languages – especially when it comes to written text.

Indeed, NMT-based software may be able to provide a relative translation, but in a business context, rough is not good enough.