When I came home yesterday from the bookstore where I buy recently published books on sale, (mostly mystery and suspense – I bought three of them this time and I already started reading “Poison Flower” by Thomas Perry – great read!), I saw that I had a message in my voicemail from somebody who was offering me “a copy editing opportunity” and urging me to call her back.
Copy editing opportunity? What is this about? I wondered. I don’t copy edit, except for editing my own translations and those of the people who work for me. I’m a translator and that means that I translate, mostly Japanese, German and French patents. I suspected that something weird was going on, but curiosity got the better of me and I called the number anyway.
It turned out that the person who called worked for a brand new translation agency that sells, or is trying to sell, post-processed machine translations to its customers. Although the message that initially caused my confusion was about a “copy editing opportunity”, the young lady wanted to see if I would be interested in working as a post-editor of machine translations for them. She said that these machine translations would be already post-processed (although she called it “edited”) by another translator, and that I would be doing the final post-processing, which she called it “copy editing”.
Copy editing of post-processed machine translation? What a strange way to put it. Not exactly an honest way either. Copy editing normally refers to editing documents, mostly for style, if these documents had been written by humans and were to be published. But how do you add correct style to something that was originally vomited up by a machine and thus may make no sense whatsoever?
I figured that they probably just didn’t want to call it “post-processing” because this particular activity has a terrible reputation among translators.
The person who left a message told me that she was very pleased that I called her back, especially since, as she put it, “I was qualified” for several languages. That kind of pleased me too, although I wasn’t quite sure how exactly she was able to determine that I was qualified for post-editing of machine pseudo-translations. Maybe because it’s not that difficult to be qualified for something like that?
“But you would be paid both for changed words and for non-changed words”
So, out of curiosity, I asked her how much would this “copy editing” pay. Well, the editor would be paid one (1) cent a word, said the young lady … but then she quickly added, “But you would be paid both for changed words and for unchanged words” to sweeten the deal. (Wow! How generous of them. They would pay me 1 penny even for the words that I only have to read and compare to the original text, even if I don’t have to change them!)
A slight pause in our communication ensued as I was quickly running calculations in my head.
When I edit my own translations or those of other translators, my typical speed is about 1,500 words an hour. Even though those are very good translations and all I have to do is look for typos, make sure that the numbers match the numbers in the original language and nothing was omitted, sometime I am slowed down by something in the translation that doesn’t seem to make sense and I have to go back to the original text and read it carefully to make sure that I understand both the original and the translation correctly.
And sometimes I still have to look things up either in online or in print dictionaries, in Google definitions or data, and look at the text of similar patents, which slows me down even more.
About 1,500 words is in my experience is a good speed when editing texts translated by highly experienced translators
If I were to edit machine-mistranslations that were post-edited by some unfortunate human who is willing to work for one cent a word (unless this poor human is paid only half a cent a word and the remuneration of the final “copy editor” is doubled in view of the importance of his task), I would probably be able to “copy-edit” about 500 words an hour to whip the text into some kind of shape that would make some kind of sense because I would have to be going back and forth all the time between the original text and the twice post-processed miracle of translation technology.
Let’s be generous and assume that I would be able to “copy-edit” about 700 words per hour. That would mean that my hourly wage would be seven dollars an hour.
After the awkward pause caused by the greedy calculations performed at breakneck speed in my head, I told the agency recruiter that I wasn’t interested in the job and I ended the conversation.
Since the Federal minimum wage in the United States is $7.25, except for states where it’s higher than that, that translator hunter was offering work to me for her outfit for less than minimum wage.
Less than minimum wage was what a company offering post-edited machine translations to clients was willing to pay an experienced translator who has a degree in Japanese studies and who has been translating patents and other technical documents for almost three decades.
The ATA (American Translators Association) Chronicle has been publishing articles celebrating the progress of post processing of machine translation for many years now. Not once has it published an article that would question the premise behind the logic and usefulness of this fatally flawed approach to intellectual activities represented in this case by translating and writing. This, to me, clearly demonstrates to what extent the ATA is controlled at this point by “the translation industry”.
To say something critical on the pages of the ATA Chronicle about the ways in which the “translation industry” is trying to use machine translation, a very useful tool for translators and non-translators alike, namely to do away with human translators as much as possible and turn them into underpaid, post-processing factotums so that some translation agencies could make a killing in this manner is simply unthinkable. The probability of something like this occurring under the current ATA leadership is about as high as the probability that we will see a North Korean citizen who doesn’t uncontrollably sob and cry during a televised funeral of the latest North Korean leader under the current leadership of North Korea. That is how perfectly “the translation industry” controls the ATA and its Chronicle at this time, which calls itself “The Voice of Interpreters and Translators”.
Only the party line is allowed for analysis of current attempts to turn translators into post processors of machine translations – ALL RESISTANCE IS FUTILE!
The only interpretation of attempts to use machine translation in order to turn translators into post-processors that is allowed by the only people who matter in the ATA (corporate translation agencies) is that post-editing and post-processing of machine translation is totally, absolutely and definitely the way to follow translation agencies into the glorious future of what used to be called translation technology, and is now called language technology. If you raise any objection at all based on your experience as a translator and simple logic, you will be immediately called “a Luddite”.
Who gives a damn about the poor human translators who are expected to work for one penny a word, possibly even less? Not the ATA, that’s for sure.
There are two articles celebrating the progress achieved in current techniques for post-editing of machine translations in the current issue of the ATA Chronicle. One of them, titled “PEMT Yourself” (which to me sounds very much like “Go and F*CK Yourself!”), begins as follows:
“It’s very common to hear translators complain about having to work on projects involving post-editing machine translation (PEMT), which tend to pay a lower per-word rate compared to human translation.”
The article is written by William Cassemiro, a Brazilian who is introduced to us as somebody who “began his career as a translator of technical manuals and who now serves on the board of directors of the Brazilian Translators and Interpreters Association.”
Having read the introductory segment, I only skimmed it quickly. Here is a sampling of the article’s subtitles:
IT’S HUMAN NATURE TO FEAR THE UNKNOWN
WILL MACHINE TRANSLATION FORCE US TO BECOME PROOFREADERS?
RESISTANCE IS FUTILE
I am guessing that Mr. Cassemiro is not working for one penny a word for translation agencies as a post-processing conscript attempting to resuscitate sentences butchered beyond any recognition by computers with algorithms. Based on the subtitles in his article, especially the last one, it’s much more likely that he oversees a stable of miserably paid indentured servants who work in this manner for him.
The second article analyzing post-processing of machine translations in a similar manner in the current Chronicle issue, called “Beyond Post-Editing: Advances in Interactive Translation Environments”, was written by Spence Green, who is introduced to translators as “a provider of interactive translation system who has a PhD in computer science from Stanford University and a BS in computer engineering from the University of Virginia.”
I’ll go out on the limb here and guess again that he is probably not one of the post-processing slaves (will they now be called copy editors?) forced to work for a penny to create something that make sense out of MT detritus. Since the introduction doesn’t say anything about his linguistic prowess, I will guess again that he doesn’t know a foreign language, at least not well enough to translate anything from or into it. You don’t need to be a translator to write for the Chronicle. You just have to be saying the kind of things that are allowed to be published by The Voice of Interpreters and Translators.
This article was actually a little bit more interesting because it analyses the history of machine translation and of human-machine interaction. But I just scanned it again as I don’t really care about things like whether “an adaptive MT product will likely bring a deeper level of MT integration to Trados”, especially when it’s presented from the viewpoint of non-translators who have a tiny conflict of interest when they are promoting and selling system based on what I consider to be a deeply flawed approach to human thinking and translation.
Machine translation can be used as a liberating tool – or as a tool to dig your own grave
Machine translation is indeed a marvelous, liberating tool. It liberates and reveals the meaning of words that used to be hidden in foreign languages for most people until machine translation appeared on the scene already several decades ago. Some of the meaning hidden in a foreign language can be revealed to people who don’t speak that language with a few clicks of the mouse by a machine translation program, generally for free.
But machine translation is not, and never really will be, a replacement for human translation. Because machine translation, or pseudo-translation if you will, always results in more or less important mistranslations, it is extremely unreliable and can only be used for materials that aren’t very important.
And sometimes it makes no sense at all, which is why “the translation industry” needs to ensure the cooperation of translators willing “to copy-edit” words that have been produced by non-thinking and non-feeling computers that basically only understand one thing: how to run calculations at a very high speed.
When “the translation industry” is telling translators that post-editing, copy-editing, or whatever else they may try to call it in the future, of machine translation is an important tool that every translator should add to his or her arsenal of tools, they are saying this because they know that without our cooperation, they won’t be able to make any money from their greedy schemes of how machine translation should be used. And the ATA Chronicle thus becomes a handy medium for “the translation industry”, has been for a number of years now.
I agree that machine translation is an excellent tool. I use this tool all the time. Unlike a couple of decades ago when all I had at my disposal were obsolete dictionaries, machine translations of patents, available for free on several websites, give me the gist of the meaning of the material that I am translating with a few mouse clicks.
But that’s all it is, and for translations that really matter, such as those on which most professional translators toil for a living, that’s all this tool ever really will be.
The reason why “it’s very common to hear translators complain about having to work on projects involving post-editing machine translation (PEMT), which tend to pay a lower per-word rate”, as the article in the current issue of the ATA Chronicle put it, is that most translators realize that the greediest and most unscrupulous sector of “the translation industry” is trying to turn this very useful tool into a tool that a certain segment of “the translation industry” is trying to use to force translators to dig their own grave with.