We will be the first to admit it – Google translate has its uses. As it currently supports over a hundred languages, getting the gist of a foreign word or phrase you encounter has never been easier. But machine translating text, with no human involvement, is actually called just that – gisting.

Word-For-Word Translation

In the early days of Google Translate, a colleague and I translated the phrase “out of sight, out of mind” into Spanish and other languages. We seriously laughed when the phrase was translated back into English as ‘invisible insanity’. There are many other hilarious translations rendered by this early technology where each word was taken literally and out of context. For instance, a clothing label that read: “Made in Turkey” was machine translated into Spanish as “Hecho en Pavo.” For those non-Spanish speakers, ‘pavo’ means turkey the animal – not Turkey the country.

Adding Context

To counteract this, Google Translate has started using longer phrases in order to impart some context. As more and more people use the system and provide feedback, it will continue to improve.

What the is not immediately obvious is that Google Translate is not actually just a very large dictionary. Rather, Google is doing what it does best, running parallel searches on a huge amount of material. For example, it can pull the text of the latest bestseller and run it against the same text in a different language. The algorithms behind the translation feature can then use mathematical probability to give context to what is being translated. If it sees that in 90% of cases the phrase [x] is rendered as [y], then [y] is what it will spit out. With ever-more (human translated) text being fed into the database, the machine-rendered translations are becoming increasingly accurate.

Tools For Professionals

Linguee is a web service that provides an online dictionary for a number of language pairs. Unlike similar services, such as LEO, Linguee incorporates a search engine that provides access to large amounts of bilingual sentence pairs found online.

SDL Trados Studio is a computer-assisted translation software suite that helps translators ensure quality and increase productivity. At Bilingual Communications, we rely on SDL Trados software for some of long-standing repeat clients in the food, energy and banking industries. This also helps ensure consistency over years and decades of work.

Where Are We Headed?

So is the future of translation one solely populated by machines? The current state of affairs is that machines are fairly good at translating, but they are not reliable yet. A significant and crucial gap remains between machine-rendered translation and that done by humans. Long sentences remain particularly challenging, and even the best AI continues to translate one sentence at a time. The predicted date for machine translating perfection was three to five years in the future – in 1954. Based on the state of affairs in 2018, we’re not quaking in our boots.

For a bit of fun with machine translation, check out this website.

Sources:

https://www.economist.com/technology-quarterly/2017-05-01/language
https://www.nytimes.com/2015/06/07/magazine/is-translation-an-art-or-a-math-problem.html
https://www.nytimes.com/2016/12/14/magazine/the-great-ai-awakening.html