The possibilities of machine translation

Is machine translation really of any practical use to professional translators? In a recent project, the Content Services team found a surprising but promising way to utilise the MT plug-in in their translation software.

The subject of machine translation is a hot topic in the translation industry at the moment. Whilst many clients are keen to give it a go, it seems to me that many translators choose to approach the subject with sarcasm, hostility or downright fear for losing their jobs in the near future. Instead of gloating every time Google Translate gets it wrong, or panicking every time it gets it right, wouldn’t it be more constructive to think about how MT can benefit translators? Let me illustrate with a real-life example.

The project

We recently received a translation request about winter fishing in Norway from German into English, Swedish, Danish, Dutch and Russian. As most editors have only a limited working knowledge of German, our established practice is that the German editor makes a first translation into English, which I edit into a text that can be used for translation by the other languages. This is clearly a rather time-consuming process, so we decided to experiment a little bit.

The experiment

With the help of the machine translation plug-in in our translation software, we simply cut out the first step and went straight to the editing stage. The question, of course, was whether the machine translation would be reliable enough to use as a base for text creation. Answer: Yes, the machine translation was good enough. It is obviously not the most sophisticated piece of writing that comes out. In fact, it is a prime example not just of particularly bad writing, but of writing full of corny mistakes. How about “biking on the bars in the sea” or “all guests at the camping is a freezer to freeze their catch available”? In context, however, it was pretty easy to figure out what it was trying to say. Whenever there was any doubt about the meaning, I would simply ask our German editor to clarify.

machine-translation-example1
First draft: machine translation

An interesting point to make is that MT was used not to aid translation into my own language (which I know better than the machine) but to enable me to work with a language I do not usually translate from. It is also worth noting that instead of making the German editor redundant, it rather freed up some time for him to get cracking with other translations – one of which, as it happened, was a Dutch to German translation, completed with a little help of machine translation in liaison with our Dutch editor.

A concern

An obvious concern is whether the use of machine translation is going to bring down the already low translation rate further and make translators bankrupt. Well – I’m actually hoping that it will have the opposite effect. Clients who are satisfied with a quick, cheap and non-edited text can be referred to the free services of Google Translate. Clients who want an effective, well-written multilingual marketing text should be charged the normal rate for copy-editing. In my opinion, the rate for translation ought to be even higher as it is clearly more difficult to produce a fluent, engaging translation than it is to write directly in your mother tongue.

machine-translation-example2
Second draft, edited by a human

Conclusion

Exciting times indeed – whilst the output of machine translation programs is currently far from publishable, it is nevertheless good enough to enable us to translate from other languages than English. At this point in time, the process may not be suitable for all types of projects, clients and language combinations. But considering the blinding speed of development in this area, I guess that it is only a matter of time before machine translation becomes a natural part of our working day, and it would be madness not to explore the possibilities this opens up.

In the near future, we will hopefully be able to let the MT program take care of the dirty work whilst we focus on the fun stuff that really matter – the editing, the style and the QA. Even if the machine translated output contained less mistakes and was more fluent and natural, it does not take into account factors such as the target audience, the client’s house style and SEO – all very important aspects of multilingual content creation. Hence translators who are worried about becoming redundant may want to start honing their editing skills.

The result

So what about the result of our experiment? Feel free to judge for yourself – it is currently live in all five languages at visitnorway.com.

This blog post is a revised version of my blog post The possibilities of machine translation, or why translators should start honing their editing skills.
Published