By Mark Heaney
When it comes to translation, anything machines can do, humans can do better. Much better.
In our wonderful whizzing world of gadgets and gizmos, you could count on one hand (or on your recently acquired Fingercounter 5000) the number of jobs that a machine can’t do anywhere near as well as a human being.
We’re still waiting for the Wallace and Gromit-style wake-you-up-dress-you-and-make-your-breakfast-in-one device – well, at least I am. If I were a sheep with thoughts of one day having a wooly family of my own, I dare say that a sheep shearing machine just wouldn’t cut it – or more than likely it would, wherein lies the problem. And for boxing aficionados, I’m sure, the sight of a scantily clad robot strutting about the ring signaling the next round would have a far from knock-out effect.
Yet perhaps one of the fields in which humans most outperform their technological counterparts is that of machine translation. Machine translation has been around since the 1950s when the powers that be in Cold War-era America decided that programmable computers and artificial intelligence were the perfect solution to the need to translate reams of Russian technical articles into English. Computer scientists duly set about the task and yet, many decades on, it is still very much a work in progress.
So, why so difficult? Well, as Alan K. Melby (in his study Why can’t a computer translate more like a person?) puts it: “Translation is difficult, even for people.” In a nutshell, languages are difficult. Language, for that matter, is difficult. It is a living breathing thing. It changes – constantly, sometimes it seems on a daily basis. Keeping up with it can be challenging. Moreover, language is laden with nuances, inflections, multiple meanings, idioms, idiosyncrasies – getting a handle on these is far from easy. Add to that the grammatical rules – and perhaps even worse, certainly for learners of the English language, the exceptions. Then you have the structural differences between languages; one of machine translation’s major downfalls is due to word order – for example, the order in German is completely different to that in English.
To top it all, you have to consider other factors beyond the realms of possibility for a computer (and sometimes mere mortals), factors such as tone, mood, context and culture. As Stephen Budianskypoints out, computers lack common sense. Unlike human beings, or at least the majority of them, they don’t live in the real world.
Add all these things together, put them into a computer, and you have a recipe for disaster. Or, depending on your perspective, comedy genius. To get an idea, let’s look at a few phrases translated into German from English and back into English using Babelfish. “When it comes to translation, anything machines can do, humans can do better” comes out as “If it comes to the translation, works on everything can do, humans better do by machine can.” OK if you want to come off as Star Wars’ Yoda with a hangover.
“Football is a game of two halves” becomes “football is a play of two halves“, which, taking the theatrics of modern-day footballers into account, is arguably not so wide of the mark. A strange chat-up line used by a Spanish friend of mine, “Would you like to have an apple with me on Wednesday?” returned the even stranger “Did you become to like an apple with me on Wednesday to eat?“. Clearly, a mixed bag (or “strewn depot”)of results.
Unfortunately, at the time of writing, the Babel fish of The Hitchiker’s Guide to the Galaxy fame (a fish that when inserted into the ear renders the wearer fluent in all languages) remains a thing of fiction. Interestingly, the Spanish talk about having to ‘change the chip’ when speaking in a foreign language, a nice metaphor for the different mindset needed to cope with the myriad rules, idioms and nuances of foreign languages. Alas, for those trying to translate into other languages, there are no such fish and chips, no quick fixes. When it comes to word-perfect translation, therefore, best leave it to the professionals.