Naturally, I don't know whether you truly are weary, but judging by the weariness of so many around me, it seems all but likely that you may be feeling the same end-of-year fatigue that I'm feeling.

There's good news, though!

It is indeed the end of the year, and for many of us that'll come with some rest. Be kind to yourself and those around you during these days!

Just one quick thing, and then I'll let you go have that well-deserved rest (though Josh, Dorothee, and Alan Melby will want you to look at what they have to say below as well!).

Throughout the last couple of weeks, there has been an interesting discussion on Twitter about what the latest developments in machine translation mean for professional translators. The discussion was sparked by an article by the Economist's Lane Greene about exactly that topic. Lane has always been interested in language and translation and, as an extension, in machine translation. His often thoughtful and well-balanced articles about our profession stand in sharp contrast to those of most other journalists who report on the same topic. If you're able to get behind the Economist's paywall (or, of course, if you're a bonified subscriber), it's a worthwhile read.

The discussion that developed on Twitter is unfortunately difficult to follow because of Twitter's aggravating inability to display discussions once they fork into different strands (something that certainly will be fixed once there's a new CEO!), but I combed through it to document the different arguments that were pontificated about (the discussion did have a tendency to do that).

I think it boiled down to essentially three arguments:

  • Post-editing of machine translation has proven itself to be a failure because it does not produce high-quality translation and contributes to the de-skilling of translators (i.e., translators who lose their translation skills).
  • The "Human-in-the-Loop" model, however, is the way forward. In the case of machine translation, this model decidedly doesn't refer to the reactive role of the translator who just fixes errors at the backend (i.e., post-editing as is typically practiced today), but to an interactive, proactive, and steering role.
  • For specialized translation (with or without the use of machine translation), a mere translation degree is not sufficient. Either an additional specialized degree in the area of the respective specialization, additional working experience in that field, or translation as a second career after developing specialized knowledge are must-haves. Translation academia has largely failed by teaching technologies of yesteryear.

Hmm. I just finished writing an article about the current state of translators (I mentioned this in a previous edition of the Tool Box Journal, and I will share it once it's published). My main conclusions were "I don't really know" and "translators are going to be okay for the time being."

There are at a minimum three million largely unaffiliated translators worldwide with a range of experiences that neither I nor anyone else has any way of measuring. I can certainly see what's happening in my professional life, and I can maybe see what's happening in a few other (possibly well-curated) professional lives. But does that enable me to make sweeping statements about where we're headed? Maybe where my particular sub-tribe is headed, but that's about it.

Still, we can see some trends for some segments in the translation sector. We know, for instance, that most large LSPs (and increasingly many smaller ones) primarily use some kind of post-editing workflow -- and, while we don't have the latest numbers, they're making pretty good money doing it. So I think it's (unfortunately) wishful thinking that the post-editing model has failed. In fact, for a certain part of the market, I'm afraid it's here to stay for a long time to come. This doesn't mean that other, more innovative ways of translating alongside machine translation won't also (continue) to exist -- and, as we've said many times before in the Tool Box Journal, we need to continue to find ever better ways and technologies that allow us to do that.

I think it would be great if most translators who have access to state-of-the art machine translation (which obviously depends on language combination, open internet access, and many other things) would be willing to upskill in learning how to train, maintain, tweak, and integrate MT. But while I don't doubt that more of us will be doing that, I sense that this group still won't be in the majority in the foreseeable future. There are many reasons for this, including the fact that while most successful professional translators have some kind of specialization, it's either not narrow enough to warrant the investment into MT training, etc., and/or there would have to be too many trained engines for the individual translator to achieve the level of distinction from generic engines that we're aiming for.

I also find the comments on the failure of academic translation programs less than helpful, considering the wide range there as well, but I do think there's good food for thought in what training for specialized translators needs to look like. It's important to remember, though, that no one is a "generic person." We all have specializations, and while some of them might be easier to match with real-market expertise than others, those existing interests/areas of expertise might well provide good springboards, jumpstarting a marketable expertise rather than looking at a virtually undoable eight or more years of education for a profession that will continue to pay less than, say, doctors and lawyers in most parts of the world.

πŸŽ„ Happy Holidays! πŸ•Ž