Faking it in foreign languages: How far can you get with Google?

On the back of reading this great article by Nataly Kelly on clearing up the top ten myths of translation, and after recently reading up on Translation and Technology, I thought I’d have a little look of my own at myth number nine, that ‘Machine translation is crushing the demand for human translation’, with a bit of research into the most popular, free, and supposedly most advanced online translation service: Google Translate.


Google Translate works by using the capabilities of its own search engine to sift through a vast corpus of hundreds of millions of documents to look for matching phrases in texts that have previously been translated in order to make a kind of educated guess at a suitable translation, and with over 200 million monthly users and 65 languages supported, it certainly gets its fair share of usage. But how much can someone with little or no knowledge of the foreign language realistically gain from this service? In order to find out, I thought I’d test it with some French texts from a wide range of genres that I have previously translated to see how effective its results are in comparison.

The first thing to note is that the translator is infinitely better now than a few years ago. Instances of struggling with the most basic phrases, as was often the case in its first few years of existence (readily acknowledged by the guys at Google), are few and far between and their heavy investment over the years has clearly been put to good use. When translating a highly specialised medical text packed with technical jargon, for example, the end-product was quite remarkable. This is no doubt due to the vast amount of material in the field available online and the ready equivalents for terms in the two languages, but it was nevertheless a very pleasant surprise.

Several of the other translations were less convincing, however, with simple slips creeping in (one such example saw ‘Echecs et succès’ [failures and success] coming out as ‘Chess and successfully’ – echecs can mean chess too but how often would you talk about chess and success over success and failure?!) and all of the translations would certainly require some level of post-editing to make them professionally usable. As such, the idea of a fully automatic and free machine translation service still seems like a very distant concept and when the extent of revision is so great that the text has to be practically retranslated, machine translation today still seems to offer little beyond the most basic of gist translations.

But this is not to detract from the service as a whole: the key idea behind machine translation is that it is fit for purpose. How often does somebody go onto the site hoping or expecting to produce a publishable text in another language? In terms of being used for more realistic and more manageable purposes such as a handy multilingual dictionary, as a way to scan texts for key information or to just get a general idea of what a text says, the system works perfectly and, as long as its very definite limitations are recognised, it remains a valuable tool.

All in all, the myth above remains exactly that; machine translation is not going to usurp human translation any time soon and is an area that should be embraced by the translation community rather than feared as an enemy looking to bury the profession.

A machine translation system such as Google Translate is undoubtedly an aid to millions of people every month, translators included, but if you’re looking to skip those language lessons and make some easy money as a Transgoogler, it might be worth waiting a few more years.


2 responses to “Faking it in foreign languages: How far can you get with Google?”

  1. […] way to becoming a global player? Networking, TermCoord glossary links and Swedish writing course Faking it in foreign languages: How far can you get with Google? Does It Makes Sense To Charge For Translations By The Word? Update on the Court Interpreter Chaos […]

  2. […] In this particular context, the word ‘corpus’ – coming from the Latin for ‘body’ (and with its lovely plural of ‘corpora’) – simply refers to a large and structured set of texts (nowadays usually electronically stored and processed) used to carry out statistical analysis and check occurrences or validate linguistic rules within a specific language territory. These days, corpora are often found at the root of machine translation technology, including Google Translate – something I’ve previously explored on my blog. […]

Leave a Reply

Fill in your details below or click an icon to log in:

WordPress.com Logo

You are commenting using your WordPress.com account. Log Out /  Change )

Facebook photo

You are commenting using your Facebook account. Log Out /  Change )

Connecting to %s

%d bloggers like this: