Start a new topic

Machine Translation: yes or no?

My statement of today is:

Blindly depending on MT engines is unacceptable, but not using valuable suggestions of several available MT engines isn't acceptable either.

Or: As a professional translator one has to take suggestions offered by MT system into account, in order to improve quality of translations.

Am I opening a can of worms with this statement?

They'll be soon out of business, Mario. There will soon be software that compares a translation with the automatically generated MT translations of the source text. (Actually this is already possible with CafeTran Espresso 2019.)

'Translators' (frauds) who score a match higher than say 30 %, will be placed on the black list.

Allow me to doubt that, please. I have been hearing this since many years now, and such translators are still around. Of course I speak based on my experience in Japan, where the bulk of my revenue comes from. I already said somewhere else: when translation companies and end users alike rely upon Google Translate to check my translations, I cannot expect they are very demanding in terms of translation quality. 

Maybe in the western world the situation is different. Some even tell me that I am allowed to use MT (although they rarely have the means to understand if I do and when I don't) provided that I correct "all syntax and grammar errors". This seems enough to make them content.

Like Jean already said: post-editing MT is horrible. Here a recent and frightening posting from Proz: Toekomst van mijn vak: I am an in-house translator at a large translation agency. I've enjoyed my job very much until now - I've always been treated very well and paid fairly. But now it seems the industry is switching over into MT, and my company is trying to keep up. I recently learned that my company plans to switch over completely to MT as much as possible, then run as many of these post-editing jobs in my language pair through me as much as possible, to ensure their quality. The idea is that my productivity will be increased so much that they will be able to give me quite a substantial raise in the near future. Here's the thing: I'm on my 4th MT post-editing job and, well...I hate it. It's not translation at all - it's correcting someone (something?) else's work. There's a reason I didn't go into proofreading or quality assurance. I like translating, the creative and research process of translating. The process of MT post-editing is something else altogether. Everything I loved about my job is basically gone. Has anyone else gone through this, or going through this? I am the sole support for my family, and I am now worried that unless I want to be a proofreader for a machine, my job is about to be phased out, sooner or later. Where to go from here? Any words of wisdom? Advice? Light at the end of the tunnel?

Sorry for being pessimist, but unfortunately there is not much we can do except sympathizing with this colleague. It’s hard to admit, but we the “traditional” translators are the big looser in the industry—at various levels of course. Even if we embraced MTPE world fully, we wouldn’t be able to sustain a family as we did until the recent past, unless we are blessed with an extraordinary amount of physical energy (not easy for people at my age!) to work much more in order to hopefully earn more or less the same, but without the motivation the translator needs. 


Then, also consider the growing number of people in the world who can read and write English, which means that more and more translators are becoming potentially superfluous every year despite translation schools and universities keep churning out thousands of new enthusiast/naïve competitors every year, everywhere. Look at you can see four new members every day in the "Welcome new members" section.


I think that the only winners in this situation are translation companies thanks to their ability to exploit the MTPE business to their advantage, Therefore, our only hope to be able to prolong the agony is finding direct clients, most of which don't even think of sending out machine-translated texts for revision—yet.

There's an interesting discussion going on here:

This particular answer struck me as a smart one:

Do you not see the difference between assessing a machine's creative flair and demonstrating your own? having to research randomly chosen terms before deciding that it's no good and you'll have to research the term yourself?

I did PEMT once, when a PM "forgot" to mention that it wasn't a human translator I was proofreading. It was totally soul-destroying; you have to first assess each sentence, pick it apart and groan at the nuances that have been overlooked and the poor choice of wording and the clumsy effect of a million different translations scrunched into something that looks like it makes sense but actually is always a bit off. When proofreading a human, you get to know their foibles and you know they tend to make certain mistakes. You know to look out for false friends. With a machine, there's no predicting the mistakes, and something can sound fine... then suddenly you realise it's not a false friend or even a true enemy but some slimy inbetweeny gibberish.

> Do you not see the difference between assessing a machine's creative flair and demonstrating your own?

In practice, I don't see the issue provided that the rate remains the same.  If you don't enjoy MTPE, just clear the segments of MT matches (e.g in CafeTran via Task > Remove target segments), and then enjoy your own creativity. Before that, you might create a TM out of MT segments (Project > Export and exchange > To TMX memory...) just for reference.

Login to post a comment