I should have written something about this a long time ago. Today, we’re already in the era where voice dictation and post-editing are old news.
Let’s start by defining post-editing: post-editors take the product of machine translation and revise it. They read through the translations provided by Google Translate and make the necessary changes to transform that sometimes bad product into something useful. So yes, the basis of post-edition is our nemesis: machine translation.
Where did post-editing come from?
With the rise of machine translation services, and in order to cut down on translation costs, some clients began asking linguists to revise machine-translated texts, instead of simply translating them from scratch.
After that, some translation software (or CAT Tools, as I describe here) started being integrated with machine translation engines. That machine translation integration proved to be something very useful for some language professionals in certain areas. Having the text machine-translated and revising it later proved to be much faster than normal translation. A larger output meant the possibility of accepting more jobs and, therefore, more income.
This spread quickly, and many translators and translation companies were pre-translating their texts using machine translation and then proofreading those translations.
Did it really mean more income?
Some translation companies noted that some translators – mostly technical translators, because that’s where machine translation works best – were able to do the same job with less effort. Their thinking was that if it requires less effort, then it’s worth less.
Soon enough, some linguists were accepting lower rates to do post-editing. From here on after, it’s the same old story: low payment generates poorer quality.
Does post-editing mean bad quality?
Now, this is a tricky question. If you deconstruct post-editing to the point that it’s almost not post-editing, then no, the quality won’t be any different.
What I mean is that you can use machine translation as a reference. You can even insert the machine translation into the segment you’re translating, and if you can keep an eye on that translation and still do your own translation and interpretation of the source text, then the translation will be human, and the machine translation will only be there to help you with difficult sentences or terms.
This may seem like a simple task, but trust me, it’s harder than it looks. There may be academic studies about this that I haven’t read, and please correct me if I’m wrong, but it’s very hard to construct a new translation for a source text of which you’ve just seen another translation. It is my opinion that the translator will always be influenced by the machine-generated target options and will struggle to pull away from them.
One could argue that, at this point, we’re not doing post-editing anymore, and that once we’re this far from the machine-generated text, we’re talking about translation again… With a helping hand from computers. The problem is that a fine line separates the two. Also, it’s very easy for a translator to be influenced by the machine generated text at any time. Therefore, we have to suppose that, taking tiredness into account: 3000 words later, it will be even harder for a human not to opt for machine-generated options (don’t get me wrong, these options can, sometimes, be quite good).
Is post-editing going away anytime soon?
It has become an area of translation, like interpreting, so I don’t believe it’s going anywhere.
Post-editing is used by most translation companies, and by many freelance translators around the world. The difference is to what extent they use it. Some use it more responsibly – only using the machine-generated translations for reference, and applying major changes – and others abuse it, keeping revision to a bare minimum, and therefore, sometimes, providing poorer quality.
With this wide degree of usage, where the bridge is being built between human and machine translation to get the best of both worlds, post-editing will not be replaced, simply because the price of purely human translation may be too high to be competitive, and the quality of machine translation is not good enough.
Is it good for the profession overall?
Right now, in my opinion, I think the only danger that post-editing brings is the information it passes onto machine translation engines. If you’re a translator and you do post-editing, have you ever considered that each and every word and translation you submit into google translate will be used by it to provide future translations to everyone else?
(Also, are you aware that, as I said before, Google keeps that data and that by using the service you agree to share it with them? Be careful, the RGPD is here…)
What I mean is that when feeding pieces of your translation into machine translation engines, you’re basically teaching them translation tricks. Thank God they are slow learners, because if they were good, we would just be translating idiomatic expressions and sentences out of context.
At the moment, I don’t think post-editing represents any danger to the profession (besides the one mentioned above). A few years ago, we were translating on paper, and post-editing is the proof that the translation market is evolving and keeping up with the advances of humanity, technology, businesses and globalization.
Should the same rates be kept?
In a world with only blue skies and sunshine, yes.
Why? Because in order to provide a top quality post-editing service, the revision must be thorough. A thorough revision of a different language requires the same skills as translation. The revisor that ventures into post-editing must know the source language vastly, be a native speaker of the target language and have linguistic knowledge of both (because there can – and most likely will – be linguistic issues with the target text).
However, bearing in mind that there is a much higher output with post-editing (because there is no need to actually type the words, but to type only some of them), it requires less effort from the linguist and therefore the rates for post-editing are lower.
If we think this through, the prices for normal proofreading are also much lower than for translation. The only problem is that, usually, machine-generated translations require more proofreading than human-generated translations…
Maybe the real issue is that post-editing shouldn’t even be a thing…
My two cents?
Back when I was a freelancer, I had machine translation integrated with my CAT tool, so the machine generated translation was there for each and every segment I translated. I read the source, read the machine generated translated and then constructed my own translation. Sometimes the machine generated option was right and no correction was necessary.
One time I went back to open an old project that I had translated a few months before. As I read through it, the translation seemed strange to me – as if it had been translated by someone else. I opened the original document to compare the two, and I could come up with much better translations for the majority of my choices. I checked a few other projects I had done before, and I found the same problem.
What had happened? I noticed that I was being influenced by the machine-generated translations and that, because of this, the quality of my final product was declining. At that time, I stopped using machine translation and doing post-editing, unless specifically required by the client. It wasn’t worth it.