Arnold Übersetzernegger in… The Translator: Rise of the Machines

Machine translation. The dreaded MT. Does it have its place? Should we see it as a threat? Or should we embrace it? Or (as is the English way) should we quietly carry on with our business, studiously ignoring it, examining our fingernails with a schoolboy level of interest, in the vain hope that nobody will acknowledge its presence and prompt us, in turn, to do the same?

In my opinion, MT has its uses. For gist understanding of basic texts with no subtext, nuance, style or irony to be replicated, machine translation can provide us with a fast, convenient (not to mention cost-effective) means to an end. Provided the end user/reader is aware of the limitations and has had their expectations managed. Yes, there are niggles and errors of varying severity, but for basic comprehension of the content of a foreign language text, it can, as a minimum, provide an insight into the meaning of a text. Let’s face it, this will be the use made of mobile telephone apps – picture the scene: you’re sat at a railway station in Minsk, looking frantically at a notice board, trying to decipher the maelstrom of squiggles and backwards Rs that are going in and out of focus in front of you. Cursing your lack of enthusiasm in your second semester of ab initio G.C.S.E. Belarusian you begin to fret that you will miss your connection and be late for a spot of supper with Binky and Portia. Then you reach for your phone. Tap, tap, a quick photo and you realise, through the message in broken but not meaningless English on your screen that you’re actually looking at a notice forbidding travel unless in possession of a transit visa. The time table is the document next to it. Panic over.

Selling machine translation raw output of a marketing text to an end client as “ready for publication” is clearly neither ethical nor without obvious repercussions. Unfortunately, we all know of someone or “someone who knows someone” who has been mis-sold language services in this manner. Tempted by the low price, fast lead time and promises of quality, the end client has taken up the “too good to be true” offer. I find myself referring again to the age-old project scope triangle analogy – there are three primary criteria, represented as the angles of a triangle: Quality, Speed and Cost. Pick which one is less important, shift it to a less-important position and therefore retriangulate accordingly. With raw MT output, the deprioritised element is clearly Quality. With post-edited machine translation, the Quality angle is moved more central to the detriment of the Speed and Cost angles. No surprises there for us seasoned industry professionals.

So, post-editing of machine translation, or PEMT for short. Opinions? I’m betting they are along the sharp-intake-of-breath-through-the-teeth lines of “no chance!”, “the client never pays enough for the time it takes me to polish the MT raw output” and “I might as well have translated the text from scratch, the time it took me!” (expletives omitted…). I think we’ve all been there at some point. The problem mainly arises from client (agencies) not realising or refusing to acknowledge the amount of effort and time required to polish an MT’d text to publication standard. Now, re-read that sentence a second time. There are three issues behind that statement:

  1. Clients requesting publication standard output should not, in my opinion, be using PEMT workflows.
  2. If a client does not require publication standard output then they should communicate this to the post-editor (fit for purpose in terms of readability).
  3. If a client informs us that the quality of the final output does not need to be of publication standard, we as post-editors should commit in terms of time and effort to that level alone.

All very well, you may concur. However, the little demon that goes by the moniker of “Professional Pride” will pipe up at this juncture, “Go on… just one more read-through for readability, tone, register, style…” and we end up delivering the job, of a quality way above that expected, and feeling the resentment emoted in the paragraph above. Of course, should you have the time and inclination, then there is absolutely nothing wrong with exceeding expectations. That should be our byword in our business. However, if the client has requested non-publication standard output, and you have agreed your purchase order on these terms, then the time and effort spent in addition polishing the text to publication standard must be borne by yourself.  So, as a business model, we’re back to that chuffing triangle again. Honestly, I should have become a Professor of Geometry.

The simple truth is this: MT will not be taking over the role of human translators within our lifetimes or those of the next few generations. It will, however, become an increasingly oft-encountered thing in our lives. There are myriad comedic examples cited all over the Internet, but we can all take solace in this entertaining clip from the guys at All Things Linguistic: