Translation Tools – A Perfect Storm for Data and Localization


Translation tool results of Beatles song
Five translations later, the Beatles will not be amused.

 

Translation tools have been the butt of many jokes and party games over the years. The results have always been ‘Hilarity ensues’, especially when you run something through several different languages.

 

English to Chinese translation tool result of "Just sit right back and you'll hear a tale, a tale of a fateful trip"

Traditional Chinese to Hungarian – Hungarian to Urdu – Urdu to Danish – Danish to Japanese

Translation tool results five translations later

The translation tool game is now as so-last-summer as Pokemon Go

 

Now, a few translation tools have begun implementing Neural Machine Translation, which means running weird stuff through several different languages actually produces a pretty decent translation. If it’s not perfect it still manages to preserve the meaning.

Related: How Will Machine Translators Change Language Learning?

Globalization: Why it’s good for human translators

This might have been a threat to human translators’ jobs, since not only can NMT handle modern English-to-other translations, it can handle sixteenth-century English as well.

Translation tool results - Belarusian to English - Juliet's lament

bad translation - instructionsIt’s not perfect, and when you’re dealing with literary purists, they’ll pick up on small details like the fact that Juliet says ‘thou wilt’ and not ‘you will’, or change the order of ‘be but’. Human post-editing for classic literature becomes much more time-consuming and expensive when you have to preserve exactly what the playwright or author wrote. It’s less expensive to just use human translators for this sort of job. DON’T mess with The Bard!

When it comes to content that is less likely to generate controversy and outrage (or ‘high dudgeon’ in the case of British literary purists), a minor difference here and there that doesn’t change the meaning of the statement will only invite gratitude that the instructions aren’t as inscrutable as they once were.

The fact is, even as machine translation tools improve and perform with greater accuracy, human translators will be needed more than ever.

Many businesses and organizations are understandably wary about translation tools that may do a less-than-stellar job. Popular wisdom instructs them to just rely on human translators when it absolutely, positively has to be translated correctly. No one wants to insult their customers with a bad website translation.

Providing a perfect translation is not much of an issue when you only have a small website with 20,000 words or so, but if you’re an eCommerce vendor with thousands of items on your existing website and hundreds or thousands added on a regular basis, not only does human translation become prohibitively expensive but it quickly becomes a project beyond their capabilities.

Related: A Babel of Asian eCommerce

It takes a human translator eight hours to translate a little over 2,500 words, and it can be done in the blink of an eye by a good translation tool. Need to translate a million words?  Twenty-five hundred words or a million, it’s all the same to a translation tool. Only the human translator can post-edit, which takes far less time than translation, unless you’re translating Shakespeare. (Just don’t go there with machine translation. Don’t do it. High dudgeon is not pretty!)

2020: A perfect storm

This year’s advances in Neural Machine Translation couldn’t have come at a better time, as data creation is exploding. We’re talking Death-Star-blows-up-Alderaan level of explosion. Humans will have created a total of 44 zettabytes of data by 2020.

To put into perspective just how much data that is, remember how huge one gigabyte was back when Windows 3.1 came on floppy disks? A one gigabyte hard drive was only customary for a really expensive file server or that shifty-looking guy down the street who ran a neighbourhood BBS offering downloadable files of questionable taste.

One zettabyte is the equivalent of 44 trillion gigabytes. In other words, that skanky guy’s über-hard drive plus 43,999,999,999,999 more, TIMES FORTY-FOUR by 2020. That’ll be emails, photos, YouTube videos of people tripping over lawn furniture and animals getting drunk on fermented fruit, songs, text messages, business documents, marketing material, tweets, boring Facebook posts about breakfast, and embarrassingly ridiculous online flame wars about any movie featuring a female heroine.

Not all of that will need to be translated of course, but as translation tools ever-evolve toward greater accuracy, the price will drop along with the need for extremely powerful hardware to produce real-time translations. That means data and content that didn’t need to be translated before will gain priority. Opening up your eCommerce marketplace to other cultures requires website localization and customer support. With better and far less expensive translation tools, online marketplaces and business websites will become multilingual as a matter of course. Website production will include multilingual capabilities within the initial design.

At the beginning of summer 2016, automated translation tools were ‘okay’ but not great. By the fall, Neural Machine Translation had transformed them and accuracy rose sharply.

If you thought you couldn’t afford localization, think again. Oh and give us your calculator for a second. We want to show you something…

 

Yappn Corp is an enhanced machine translation company still trying to wrap its collective head around 44 zettabytes of data and how many blinks of an eyeball it’ll take to translate it all into 67 languages. For more information please contact sales@yappn.com or call us at +1.905.763.3510 x246.

 

Never miss out on another Yappn blog post! We love language, translation, and the cool machine translation technology driving it. Not to mention eCommerce, Asian growth, and how we’re all learning to communicate better with each other. Sign up for our new post email notifications today!

 

 

Written by Nicole Chardenet, Sales Development Rep at Yappn

 

Share This Post