By Bridget Hylak, LTD Administrator
Note: This article is the first in a new series, to be published in the ATA Chronicle and in this blog, created by ATA’s Language Technology Division. This series will discuss machine translation (MT) and machine translation post-editing (MTPE) from several angles. It will attempt to illustrate the promise, the pain, and the practicality of these technologies and what the present and the future may hold. Should sober linguists adapt or hold their ground? (And if either, when, where, how, and for how much…?)
Linguists old and new are celebrating and/or struggling with the influx of project requests requiring MT or MTPE. Like it or hate it, neither one is going anywhere. Mind you, though, they are two very different things.
MT
MT, or machine translation, is an undeniable and oh-so-handy tool for linguists to have in their professional toolbox that can be beneficial in many creative ways. (See sidebar.) Depending on the language combination and specialty you work in, and the specific engine you choose (there are many, so research optimal engines for your language combo), MT can become your new favorite, furry friend.
If you haven’t tried it yet or haven’t securely incorporated it into your professional work environment, reach out already and take the olive branch. MT won’t bite or upset your flow; it will purr like a kitty and lick your feet, well, also like a kitty.
If pets aren’t your thing, just think of your first car—smelling the smells, caressing the steering wheel, turning up the music, and feeling an undeniable sense of newfound freedom. In other words, if you haven’t quite cozied up to MT in your daily grind yet, where have you been all my life? Give me a call (WhatsApp also works), we need to talk. Yep, it’s that good.
While some less common, lower-resource languages are still in the process of building truly useful engines, don’t worry. The world’s getting smaller by the minute and MT is only getting better at learning—your day will come. To anyone in that situation, you have my heartfelt apologies that the powers that be haven’t yet done their due diligence. In the meantime, keep playing around with it until you start to notice, “Hey, that’s not bad….”
MTPE
MTPE, or machine translation post-editing, on the other hand, isn’t exactly the same thing and shouldn’t (yet) be discussed in the same Zoom webinar.
When it comes to MTPE, my current philosophy (which I reserve the right to update as this fast-paced topic evolves!) is simple: MTPE nQsf. Huh? That’s, “MTPE, not QUITE so fast….” Think of MTPE this way:
MT = machine
PE = you
Successful, highly accurate MTPE requires special handling by trained linguists who, at first, might find themselves completely flabbergasted by the task. MTPE is not like regular “E” (i.e., editing the work of a trained, competent colleague). Nor is MTPE anything like “T” (i.e., translation with MT in the background). Neither does it represent merely a linguistic challenge, but a technical one as well.
But wait! You didn’t become a translator to use technology, you say? You didn’t spend all those years mastering language XYZ only to have a machine cramp your style? Tell that to the neurosurgeon currently conducting robotic brain surgeries, or the auto mechanic who hooks incoming cars up to the hard drive to diagnose and resolve issues. Nowadays, and even “way-back-then-a-days,” both professionals need not only industry smarts but tech prowess as well. That is, if they want to keep their jobs.
My favorite analogy, which erupted over my own frustration with communicating both the promise and the pain of language technologies like MTPE (or even CAT tools, for that matter), is a surgical robot. I’ve shared this illustration on countless occasions with linguist, client, and government audiences, and the silence that immediately falls makes it clear that a loud, clangy bell has rung somewhere in their heads.
It goes like this. If you’re on the operating table about to undergo a right radical nephrectomy (“the old-fashioned way”) and the pre-op nurse casually mentions your surgeon will be using a robot—whoa, Nellie! How would you respond? Personally, I would jump off the table and demand to know five things:
- Is my same, excellent Harvard-trained surgeon going to do my surgery?
- Is that “robot” any good? I mean, how many nephrectomies does it have under its hood?
- Is my same, excellent Harvard-trained surgeon any good at doing surgery via a robot? That is, how much training does she have “on the bridge” between her medical degree and proper certification or “whatever” she needs to be qualified to use that thing on me?
- How many right, radical robotic nephrectomies has she actually done this way?
And finally:
- Don’t tell me that sassy, know-it-all tech guy sitting there is going to do this, ‘cuz he doesn’t even know where my kidney is.
Yes, family and friends, MTPE is its own thing, with distinct advantages, disadvantages, quirks and benefits, headaches, heartbreaks, and absolute moments of ecstasy—but you must take it seriously.
Now for the Necessary Pause…
For those hesitant about MTPE, you’re in good company. Some of my most respected colleagues assure me that they’re just fine without it, but these are mainly veterans who have spent decades honing their own niche and client base. For the rest of us, this sarcastic essay is a clarion call to think proactively instead of just drinking the MTPE Kool-Aid without reserve.
T&I veterans whose bottom line may be suffering realize that we must train ourselves to think, talk, and educate others about MT not just in conjunction with PE, but in new ways that include MT engine training, asset management, language learning, artificial intelligence, and many other possibilities yet to be discovered/proven. In fact, it’s incumbent upon us as industry professionals to lead with enthusiasm but also with care, and to keep the playing field and our minds open to the many and powerful ways these technologies can assist, enhance, and even improve our efforts and our income.
Return on investment (ROI), bottom lines, and time savings may be on the tip of every MTPE evangelist’s tongue, but as language connoisseurs who understand the power of each properly or poorly translated word, we’re obligated to gauge the success/failure of MTPE by the equally essential measures of quality, nuance, connection, and more—all of which ultimately contribute to ROI! Critical outside factors such as privacy and data protection haven’t even been addressed here, but they need to be intimately understood and communicated to high-stakes clients for whom such matters are a deal breaker and, in fact, eliminate MTPE (and other technologies such as cloud-based MT) as an option.
For the foreseeable future, many (many) jobs will be done in post-editing mode, but many (many) others will utilize MT differently. How so, you ask?
The answer is as complicated as it is simple. Professional linguists, like all professionals, need to continuously research, train, upgrade, and self-motivate to forge their own paths. Neither MT or MTPE is going anywhere, and, most importantly, neither one will threaten your career. They will only enhance it. (Unless, of course, if you ignore them.)
Buckle up and stay tuned. This carnival ride is just beginning. Until next time…