Mental trips

I was just thinking on my way to the off-license. People are getting obsessed with this topic, if machine translation will ever replace the human translator…Well, if these people just would know anything about the state of art in machine translation, they would have no fear.

My last job was to build a social network, in which the posts get translated into 3 other language in real time (immediately post-edited by humans, since otherwise it wouldn’t be understandable at all in most of the times), using different approaches of machine translation: we worked with only rule-based engines, a mixed rule-based and statistics based one and with an only statistics based approach. The results were good.

Good, if you understand under “good”: the translated text appears in the required language.

It doesn’t say anything about if the translated text makes sense, if the terminology is the correct one or if any human would understand the translated text. Well…I can tell you, we are far away from a real intelligence and it will never happen in a broad context. Google – which we all hate because of its quality – is coming quite near to be the best among the statistics based approaches, if we try it with random languages on random context (on narrow domains with large input you best build your own engine with any MOSES solution).

Sure, if they have not enough resources to guess the best option of which word has to stay close to another one, who does? But did you ever try to translate anything from or into Hungarian? Don’t expect to understand it!

Although in the translation business people are not talking about anything else more, than post-edition, every CAT tool is expected to include machine translation engines to make translator’s lives easier, it might be a bit too early to ask translators to lower their rate, because it was already pre-translated with a machine.

Should we refuse to lower our rate, if the machine translation is not good enough? Sure, we should, but will we?

Why are agencies and clients so keen to lower the rate of the person, who is making is possible for them to get bigger their business. I bet, no company would ever make such a huge success, like they are doing it now, if they should keep being local. And who will let you understand, what stands in the manuals? Do you think, you can force the fabrics to hire operators, who work with a machine, that speak at least 2 languages?

Let’s face the truth, operators won’t learn a language, if they want to work as a welder, because if they would, they wouldn’t want to work as a welder. (I love this!) So, you as client: try to grow and become internationally famous and important, so you want to internationalize your business, you set up some manuals coming with your machines, you set up a website, try to get new contracts maybe every through the internet, but you want to pay the people, who are enabling it exactly NO THING.  (there is a well-known translation agency, who is recruiting interpreters for 10 £/hour….well, I got paid this much as a housekeeper, when I was a student, sorry babe, there is something wrong in your calculations…)

And now that these clients / agencies can’t say anything more to get the translators lower their rate, they come up with this machine translation, which should then lower the translation rate. I have a vision of this: this business could work, but only if: the machine translated text is proven to be understandable or is evaluated using a serious metrics – not the BLEU, come on! – to be useful and not only counterproductive during the translation process in the language pair.  


Leave a Reply

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

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

Google+ photo

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

Twitter picture

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

Facebook photo

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


Connecting to %s