Machine translation – or simply MT – is a fantastic tool. It allows us to understand texts in languages from every corner of the globe, and (in theory, at least) enables us to communicate with anyone at all. The technology is coming on in leaps and bounds, and the next generation will grow up in a world where the language barriers are considerably fewer and lower.
As MT has made inroads into our lives, misunderstandings have also crept in – and some people have very strange ideas about MT. In this post, I will be listing some of the most common misconceptions and explaining the truth behind the rumours.
Misconception 1: MT is a new technology
MT has existed as a phenomenon since the mid-1950s. Back then, the US Army didn’t have enough personnel who could read Russian. They wanted to use computers to translate the huge quantities of Russian-language intelligence they came across every day, and so officials turned to IBM for help. IBM is said to have claimed at the time that “we’ll have perfected machine translation within five years”. Sixty years later, IBM may want to eat their words…
Misconception 2: MT will replace human translation
MT will never completely replace people for the simple reason that machines will always make mistakes. To get a really good translation, a human has to post-edit the proposed machine translation.
Misconception 3: Translators don’t use MT in a professional capacity
You might be surprised at how many translators actually use MT as a tool. It’s so common that I would say MT is already part of most translators’ workflows. Soon, translators around the world who don’t use MT will be in the minority. In fact, this may already be the case.
Misconception 4: The MT engine I linked to my CAT tool gives such poor suggestions that it takes longer to correct them than to type in my own translations
In a way, this particular ‘misconception’ may actually be true. It can certainly be the case that a translation engine isn’t ‘good’ enough and that correcting the suggested translation is a time-consuming process. As soon as a translator sees that the machine-translated sentence isn’t any good, the sentence should be deleted from the CAT tool and the translator should come up with their own translation. As a rule of thumb, if it takes longer than two seconds to correct the sentence, you should scrap it and translate it yourself. Translators soon learn to recognise from a quick glance which sentences are worth re-writing and which should be ditched. And there’s no need to take a delicate approach – computers don’t take offence if you ignore their suggestion!
Misconception 5: Google saves all the text it translates, so my company can’t allow translators to use Google Translate as a tool
The customer is always right, of course, and so we do what the customer asks. If our customer doesn’t want to let us use Google Translate for their translations, we never allow it to be used. And that’s that. But the reasoning above shows a misunderstanding of how MT actually works.
This is one of the biggest misconceptions about the world’s most popular translation engine, Google Translate, and unfortunately the rumours have become an accepted truth – but it’s incorrect. Google guarantees unambiguously that they don’t save any machine translated texts for longer than the time taken by its servers to translate them. Google simply isn’t interested in the texts translated. See for yourself under “Data Confidentiality”.
As well as these common misconceptions, there are plenty of other misunderstandings. Hopefully five of them have now been dispelled.
Whatever the case, MT is here to stay. And MT may well have been used when producing the text in your computer’s operating system – whether you’re using a Mac or a PC – as well as descriptions in Amazon listings, the handbook for your car, and so on. The list is almost endless.