Just as CAT became the new translation practice some 25 years ago, we are now quickly transitioning into PEMT, post-editing of machine translation output.
The acceptance of CAT was by no means easy for some, who saw it as an insult to their professional self-understanding. It involved recycling our own and other translators’ work rather than starting from a blank sheet. The tool provides matches of varying degrees, and it can do fragment assembly, but it does not generate its own sentences. Over time, translators got used to this and now most of them refuse to work without CAT. With CAT, rightly or wrongly, they feel in charge, while happily relying on the assistance of the computer.
With AI-supported MT and PEMT taking centre-stage, the perception has radically changed: translators now see themselves as assisting the machine. And that just doesn’t feel right.
In my view, it is this change of perception that is the biggest hurdle to accepting this new working method.
In a study described in a little more detail in my upcoming Long blog about PEMT, project managers attribute the reluctance of translators against post-editing to two factors: PEMT is thought to be “boring” and “financially less rewarding”.
As someone who first worked with Systran output some 30 years ago and now regularly uses DeepL I would say there is also a “historical” factor in the sense that historically, MT output was pretty much a disaster. And that perception lingers, both amongst translation professionals and the public at large. In the early days and right up to 2016, i.e. before AI and neuronal technology were introduced, MT was more of a stumbling block and an irritation than a help, making post-editing a very frustrating and boring task.
But times have changed, and if you use a sensible engine and get a grip on client terminology, you can actually enjoy PEMT as an activity. That’s certainly my experience.
I believe it is important to prove to ourselves and to our clients that we are doing a valuable job, that PEMT is highly skillful work that requires quick decision-making, commitment and linguistic and cultural knowledge – and doesn’t just involve “pressing a button”. Studies have shown that the end results of a PEMT process can be better than pure human translation. I would agree with that. Because MT takes some of the groundwork and tedium away and lets us concentrate on what matters! But we can only be convincing if we embrace PEMT and truly make the best of it. As a company, and as individual translators, we need to get to grips with tweaking our working methods to fit the new reality, while upholding our idea of excellence. That’s not boring.
*Confusingly, both are used, MTPE and PEMT, referring to the same thing: Post-Editing of Machine Translation