Monday, November 29, 2010


Part 3

By Bernie Bierman

CAT tools themselves are not to be regarded as our enemies.  They didn’t do anything.  But the supposed translation specialists did it to themselves and the “real” translators let them.”

The above comment made in response to Part 2 of this series certainly has merit, and I would be hard-put to dispute it.  Indeed, I would say that in respect of the economic impact of computer-assisted translation, the CAT operators or workers (né[e] translators) are not the sole source of the problem, although their complicit behavior has certainly been a contributing factor.  From where I sit or stand and from my keen observations, it has been the language service providers (more commonly known as translation agencies) who have used CAT in an exploitative manner and at the same time have converted translation from a communications service into a commodities business, i.e., the selling of words.

The ability of the various CAT programs currently on the market to undertake “precise” computations consonant with their own (computational) parameters and formulas as to how many “new” words, “repeated” words, matches of phrases or passages or segments, percentages of matches, and so on and so forth down a whole line of esoteric calculations, has resulted in among other things the formulation of a rigid pricing structure that one would commonly find in the sale of goods or commodities.  But given the particular structure of the translation industry, the biggest impact of this pricing rigidity has fallen upon the so-called “low man or woman on the totem pole”, i.e., the independent CAT worker (who has manifested over and over again his or her willingness to accept such rigidity and consequent commodification of the service).

About two years ago, a CAT operator née translator named Ronnie McKee decided to take a much closer look at the computational methods used for determining what a translation agency would pay her for her labors.  She wrote, I was asked (by a translation agency in Spain) to translate a large document, namely, a 20,000- word corporate governance report, with a TM (translation memory) provided.  I had heard about rate scales having to do with number of repetitions, etc. etc., but I never really understood what it meant.  Well, now I do.  And for those who are interested, I'm going to share:”

And Ms. McKee did share, as others have subsequently done, and her findings evidence beyond a shadow of doubt that this intellectual service known as translation has become nothing more than the sale of commodity-like units, which see:
* * * * *
The following table represents the analysis undertaken by the Trados software of the document which Ms. McKee was asked to translate.  The data shown in this table has a direct relationship to the data shown in the immediately-following table, for that table shows what the translation agency will pay the CAT operator when it provides the latter with translation memory (TM) .
Match Types

Context TM



95% - 99%         

85% - 94%         

 75% - 84%         

 50% - 74%          

 No Match          


In the case of the translation agency involved in this transaction, a segment is defined as a sentence; it is what the system joins together into one translation unit.  In line with that, the subject document for which a translation was ordered from Ms. McKee contained (according to the computation undertaken by the agency’s Trados software) 149 repeated segments, and those segments included altogether 415 words.  The 100% line means that there were 909 segments in the source-language (SL) document  that have an identical match in the translation memory   The next group indicates that 67 segments have 95% to 99% matches. Accordingly, this agency determined that it would pay the CAT worker a different rate or amount based upon the degree of the match.
The subject SL document contained 284 segments of “brand new translation” (i.e., “No Match”), and those segments constituted 6055 words.  Therefore, the agency paid the CAT worker  the highest rate for those and paid the lowest rate for the 100% matches.
The table below was used by the translation agency for determining how much it would pay Ms. McKee for her work based upon the data embodied in the above-indicated table:
Match Types

* * * * *
A cursory glance at these figures indicates that for the entire project of 19,818 source language words, Ms. McKee was paid a total amount of € 701.96, which we’ll round off to €702.00, or approximately US$990.00.  That breaks down to €0.036 per SL word or US$0.05 per SL word. 

It would be tempting at this point to undertake a full, in-depth analysis of all of these figures to determine such things as how much Ms. McKee made on an hourly basis, how much efficiency did she achieve by using robotic translation tools, along with numerous other compensation-related aspects.  However, that is not the purpose behind the writing of this article, although it would certainly make good fodder for a future article.

Rather, the purpose behind the writing of this article is to point up not just the commodification of translation, but its total and complete commodification.  But not only does this total and complete commodification elicit comments (especially in the compensation-related area), but also raises numerous questions, questions that both translators and CAT workers have raised and continue to raise, such as how the particular CAT program (Trados, WordFast, DejaVu, et al) determines what is a 72% match or a 48% or a 12% match or a 91% match, and whether there isn’t some kind of interpretive element at work in such determination.  Many CAT operators have expressed and continue to express puzzlement over such determinations (made not by them, but by the translation agency), and even more puzzlement over the methods employed for arriving at such determinations. 

And another equally striking question is how a computer or computer program can determine that two segments in a source language document are identical without having a clue as to whether the target language will demand a somewhat different or very different translation because of that very naughty word context.  Admittedly, however, this is drifting slightly from the main thrust of this article, which to repeat is the commodification of translation.

The pinpoint breakdowns and classifications indicated in the tables shown above demonstrate beyond any discussion and beyond any doubt that the communications service once known as translation has become nothing more than a robotic selling of words…with certain words being the most expensive, other words being a little less expensive, and some words being bargain-basement cheap or in some cases given away for nothing (or if one prefers, given away in the most charitable manner).

And to support and otherwise propagandize this commodification, CAT workers and CAT operators are told that they will attain heights of efficiency and productivity never dreamt of just ten short years ago, and that this efficiency and productivity will double, triple or even quadruple their earnings.  How this will come about in the context of ever-falling translation prices and rates, of more and more competition and of more and more “memorization” of vocabulary, terminology, phraseology, etc., is never explained. 

Not surprisingly, the world community of CAT workers and operators has cheerfully accepted this commodification of their work (perhaps because many never knew better and/or never knew what preceded this commodification) and, in many cases, they have joined their translation agency clients to become the champions, the salespersons and propagandists of Words for Sale.

Yes, indeed, today one can easily tell when CAT workers and operators are gathered in one of their very numerous periodic meetings or conventions or conferences: there is a long line of chauffer-driven limousines, a small but obvious congregation of pilots and flight attendants for the numerous private jets waiting at the nearby airports, and of course a parade of the latest fashions in dress that warms the hearts and pocketbooks of the most elite designers of Paris and Milan

* * * * *

Postscript:  Some time in late January or early February of every year, I prepare financial data for my personal accountant, Smiley.  The data is neatly laid out so that Smiley does not unnecessarily have to call me for any explanations.  Now, I know and Smiley knows that I know that he does not sit down at his desk and spend literally hours upon hours with an adding machine and calculator filling out all sorts and manner of tax forms and declarations (as my late brother-in-law, also an accountant, had to do for many years). 

Nay, I know and Smiley knows that I know that he takes all of the data I provide him with, inputs it into a computer and in a matter of minutes, yes minutes, out come all of the completed tax returns and declarations.  However, when Smiley sends his bill, it does not reflect 12 minutes or 22.4 minutes, or even 1 hour and 7 minutes of work.  So you, the CAT worker out there who might be reading this article, may want to ask “Then what precisely does Smiley’s bill reflect?”

OK, I’ll tell you what it reflects: It reflects his knowledge, his skills, his expertise, his unique and singular talents, and I am more than happy to pay him for that knowledge, skill, expertise and talent, It provides me with a sound sleep at night.   I trust, or at least hope that the readers of this piece (whether they are CAT workers or translators or even translation agency executives) will get my drift, if not the message.


I don’t use CAT tools, but nevertheless I feel that it isn’t fair for a translator to charge for passages that are repeated in a document.  If I can cut and paste a passage, then I’ll charge only once for the words in that passage”
   - Statement by Eve Hecht, a translator and instructor in German legal translation at New York University, made at
       the 2010 convention of the American Translators Association.

If there is any one statement among literally hundreds of thousands of statements by translators (and their 21st century successors, CAT workers or CAT operators) that so perfectly reflects their economic attitudes, it is the one by Eve Hecht quoted above.  Unfortunately and indeed most regrettably (at least unfortunately and regrettably as I see the world of translation), Ms. Hecht’s beliefs and attitudes are not just widely accepted, they have become unquestioned conventional wisdom.  In this past decade more translators and their CAT worker successors have come to believe and accept the attitude expressed by Ms. Hecht.

I have said “unfortunately” and “most regrettably” only out of politeness and what I with some hesitation would call “good taste”.  But I hope that the reader will allow me the indulgence of setting aside the politeness and “good taste”, and allow me to express myself in the terms that would be more deserving of Ms. Hecht’s economic or commercial viewpoints.

Ms. Hecht’s statement makes me want to vomit!  It makes me want to scream. It makes me want to yell out, “Is that all that you think of yourself, of your unique skills (as a translator), of your talents?  Is that what you think of yourself and the work that you do…nothing more that some clerk-like endeavor, nothing more than an extension of the clerical furniture?  Is that all you think of the value of your time?  Is that all the self-worth you have?  Is this what you teach your students, namely that translation is nothing more than typing words, “NEW” words? 

You say, Ms. Hecht, that “it isn’t fair for a translator to charge for passages that are repeated in a document”.  It is not fair to whom?  Surprise me, Ms. Hecht, and tell me precisely why it is not “fair” for a translator to charge for a “repeated passage”.  Surprise me, Madame, and defend your position.  You claim that you are a translator of legal texts.  OK.  Suppose you have a passage like this:

Pursuant to the provisions of §12, section b.2 of Article 719 and §23, section d.1 of Article 793 of the Civil Procedure Code, as amended, the plaintiff can claim…” and two paragraphs later, the text reads, “But in the case before us, the provisions of §12, section b.2 of Article 719 and §23, section d.1 of Article 793 of the Civil Procedure Code, as amended, do not apply to a plaintiff who is unable…

As I understand it, Madame, the second time that the passage the provisions of §12, section b.2 of Article 719 and §23, section d.1 of Article 793 of the Civil Procedure Code, as amendedappears, you do not charge for the translating of that passage.  According to what you say, even if this passage appears ten times within the text, it is a “freebe” to the client, because as I understand your thinking, all you did was “cut and paste” or make a macro, and that in and of itself has zero value…to you (and to all other translation practitioners whom you believe should follow your “benevolent” and “self-sacrificing” attitude).

Tell me, Ms. Hecht, if the words “pursuant to” cropped up 75 times in a text, would you also eliminate those words 74 times from your word-count and attendant invoice? After all, according to you, the only thing you did was to make a little macro and push a button 74 times, and that certainly is not worth anything. Not a single Euro cent.  No a single penny.  Not a single centavo.  Not one single Turkish Lira.   Right?  After all, Ms. Hecht,  according to what you state and believe, it would not be fair for a translator to charge for pushing a macro key 74 times, or 174 times or 374 times.  Indeed,  Madame, why charge for words like “and”, “but”, “in”, “out”, “from”, “to”, “you”, “they”, “he”, “she”, etc., etc., etc., that repeat over and over and over again within any text?

If indeed, Madame, this is in fact what you believe and what you implement in your work, then I have absolutely no hesitation in expressing my opinion that you are precisely one of the many who have wittingly or unwittingly urinated upon the profession of translation, loudly proclaiming that it has little value and what little value it has is about on a par with the most menial clerical endeavor. 

You, Madame, in my most considered opinion are precisely why the translation-buying and translation-using public views and continues to view translators as nothing more than little language clerks and extensions of the clerical furniture.  Yes, Ms. Hecht, you truly want to make me vomit.  Now, really and truly surprise me and defend your position.

* * * * *

Tuesday, November 16, 2010

ATA'S DENVER GATHERING: Some Personal Observations and Comments

The American Translators Association held its 51st Annual Convention from October 27 to 30 in Denver, Colorado.  Although attendance was nowhere near the levels that had been attained for the 50th Annual Convention held last year in New York City, the attendance by any objective measure was respectable and evidenced the association's continued success in staging these events, and this, despite the persistence of the recession that has gripped the United States since 2008, and the ever-increasing costs of attending these annual meetings (some observers place the cost per individual at close to US$2000, a cost that would cover registration fees, 4 to 5 days of lodging and meals, and round-trip transportation to the convention site).  But even with attendance down from last year, the ATA public relations machine (which now operates 24/7) immediately sent out a release stating that the 2010 convention was one of the best-attended and most successful in the organization's history.

The ATA is arguably one of the largest - if not the largest - translator organization(s) in the world, and if it is not the largest, then it has certainly been one of the most successful from a  financial and public relations point of view.  With a permanent professional staff housed in a suite of offices in a Washington, DC suburb (Alexandria, Virginia), it has forged partnerships with numerous suppliers of translation hardware, software and related technology, partnerships that have added to its coffers, but at the same time engendered some hesitating criticism from outside (but always muffled or totally silenced by the organization, which carefully and painstakingly guards its image)  

In the past decade or decade-and-a-half, the American Translators Association has moved inexorably towards a definitive technological orientation, and its 51st Annual Convention in Denver evidenced that beyond any doubt.  It might have been sheer coincidence, but ATA's 51st annual gathering immediately preceded the ninth biennial convention of the Association for Machine Translation in the Americas (AMTA), an organization which as recently as the mid-1990's was viewed as some kind of leper to the rank-and-file of ATA membership.  Today, AMTA's respectability is no longer questioned and the fact that it chose to hold its party immediately following that of the ATA, and in the very same city, says something or gives some insight into the direction that both organizations appear to be taking.

Without doubt, the attendees to the 51st Annual Convention of the ATA had more than a sufficient quantity of translation technology on which to feast.  A few of the attendees, but certainly not an army of them with whom this correspondent spoke, pointed out if not complained about the absence of attention to economic issues.  Not one of the 18 pre-conference seminars or the 167 sessions was devoted to economic issues.  Curiously, despite the ATA's obvious ignoring of or perhaps total disinterest in (or maybe just plain fear of) economic issues, private conversations among attendees in the hallways, bars and restaurants seemed more often than not to focus on economic matters.

Following is a kind of summary tour of some of the more highlighted (or perhaps more lowlighted) sessions that were offered up on the menu of ATA's 51st Convention in Denver last month:

First on each day’s schedule, and competing with the regular sessions, was a new “event” coordinated by Naomi J. Sutcliffe de Moraes, the head of ATA’s Translation and Computers Committee.  This “event”, which appeared under the heading “Learn”, consisted of a series of “tool tutorials” and took place in a special area of the exhibit hall that had been set aside for the purpose.  Each company was responsible for its own presentation, which to be expected was designed to promote that company’s products, some of which were cloud-based or SaaS (Software as a Service).  In reality, calling these sessions (at least the ones I attended) “tool tutorials” was a big stretch of the imagination. 

ATA also added a new Session Code dubbed “TIP” (Translation & Interpreting Professions).  This new code was given to “sessions that explore developments affecting the Translating and Interpreting Professions as a whole (my emphasis).  Although I was able to attend only two sessions near the end of the conference, judging from the titles of all the sessions under this code, the single most important development affecting the translation profession is the reversal of the letters TM (translation memory) which results in MT (machine translation) and its acceptance by the profession(s).  I was struck here by the similarity to the hand-wringing that never ceases in the interpreting industry, where major issues are put on a siding to allow endless discussion on whether interpreters should be called or can be called translators or demanding that the public stop calling interpreters translators (or vice-versa). 

To give readers some insight into the eight TIP sessions, I shall list the titles of a couple of the sessions I was unable to attend along with a brief biography of the speaker(s) and a few comments before giving a more detailed report on the two sessions I did attend.  Descriptive details in quotes were provided by the presenters.

TIP-1:  A Futuristic View of the Translation Profession.  The speaker was Renato S. Beninatto who, according to Milengo, Inc.’s webpage (he is their current CEO), is “a corporate strategist and market research evangelist [sic] with nearly 30 years of executive-level leadership in the localization industry.”  He “has forged a reputation for visionary leadership, most recently as the co-founder and former Chief Connector [sic] of Common Sense Advisory, the industry’s foremost market research firm.”   Notably lacking in the speaker’s background is any experience as a translator or linguist.  If I was somewhat dubious about the man’s “visionary leadership”, I was very much put off by the description of a “market research evangelist”, for I found the religious overtones a little scary.  However, since I did not see his presentation, I cannot comment in any way on his “futuristic view” or his view on what the future holds.

TIP -3:  Machine Translation:  Friend or Foe?  The speaker was Adriana Beaton, an account director at SDL International.  Her biography states that she has worked in localization since the early 90s and that “in her long career (I’m not sure if 20 years really qualifies as a “long career”, however…) she has covered [sic] many roles as linguist, project manager, program manager and globalization analyst, working for multinational corporations like Hewlett Packard, Sage Software, and Recruitmax, as well as a few localization providers.  This has given her a deep perspective on the issues that companies face every day when trying to establish or consolidate a global presence.”  There was no mention at all of any insight into issues that translators, some with more thirty, forty and even fifty years of experience, are facing.   Other than that, an internet search indicated that she was actively “tweeting” during the ATA gathering.  Again, since I couldn’t attend the presentation, I was left wondering about the question in the title and more so the answer: Is MT a friend or foe…and to whom?

TIP-6:  Machine Translation Post-Editing and Machine Translation for Productivity.  Speakers were Laurie M. Gerber, who is currently the treasurer of the Association for Machine Translation in the Americas; Walter K. Hartmann (no relation to Nicholas Hartmann, current ATA president), a translator who has been working with MT and post-editing since 1985; and Maria F. Lozano, a scientific and technical translator from Buenos Aires, Argentina, who is currently employed as the in-house Spanish translator and reviser at the Pan American Health Organization in Washington D.C.  There was a great deal of interest in this session and it played to a standing-room only crowd in a room that I estimated could accommodate 250-300 people.  I spoke with a couple of translators seated near me, both of whom expressed the opinion that, like it or not, post-editing machine translations is something that 21st-century translators must consider.  Ms. Gerber served as panel moderator and spoke briefly about the AMTA and how attitudes toward machine translation have changed are continuing to change rapidly.  

Mr. Hartmann told the audience that he began post-editing machine-translated technical literature from a large manufacturer of copy machines and then expanded into applying MT to other customers’ technical documentation, mostly in the automotive and electronics domains.  While there was improvement after 2000, source-text quality remains low and, not surprisingly, certain types of texts are more suitable than others for the MT process.  Formatting can also be a hindrance as can in-line codes, bulleted text, proper names, abbreviations and punctuation.  Nonetheless, he stated that 40% of his work now consists of post-editing machine translations and that he “enjoys the work”.  

Ms. Lozano gave an in-depth report on how the Pan American Health Organization uses machine translation.  The PAHO program was developed in close contact with human translators who give ongoing support.  Computational linguists work to improve the program and translators cooperate.  There is always a human component which corrects output and makes certain that the translation fits the target audience as well as manually checks dictionary entries.  She stressed that complete automation would never be possible and that there is still reliance on feedback from human translators.

PAHO uses two main editing modes, minimal and extensive, and translators are free to change as much as they like.  She listed some of the advantages of post-editing work such as never having to face a blank page, having to do less research (preferred terms and official country names are used; in addition, she cited the accuracy of figures, the reliability of dictionaries, the absence of omissions (the program translates everything), and the presence of uniformity (the program helps maintain terminology consistency).  “The program works well with WordFast,” Lozano said, “and is great for translating large quantities of material.”

The disadvantages all had to do with diminishing translation skills, including the loss of creativity, forgetting terms and solutions, overlooking errors because MT output “sounds correct”, and settling for less because it’s easier.  CAT operators and those working with translation memories seem to be experiencing the same things.  (There will be more on that subject in a future article.)  
While Ms. Lozano painted a rosy picture of how machine translation works at PAHO, one was left without the certainty that this rosy picture could be duplicated in other translation domains and therein lies another issue.

TIP-8:  Man vs. Machine:  Do Translators Need to Pick a Side?  This session, one of the last, was also listed under “Learn”, but not as a “tools tutorial”.  It was moved to the meeting room that had been used for the Annual Meeting and even so, it was SRO.  The description began:  “Some of the finest minds in our profession will discuss the current status of machine translation, the continued importance of human translation, and new ways the two approaches are being combined.”  The “fine minds” referred to in the description belonged to: Mike Dillinger, a past president and current vice-president of the AMTA, who has developed machine translation systems at several companies; Chris Durban, a freelance translator and recipient of ATA’s Gode Medal in 2001, who specializes in financial translation and whose claim to fame and qualification for the Gode Medal is the relentless exposure of “bad translation” and a messianic quest to make the world safe and secure from it;  Jiri Stejskal, a translation agency owner and a past president of ATA, who also chairs the Status Committee of the International Federation of Translators (FIT); and Joap van der Meer, director of TAUS and the TAUS Data Association (TDA), a global language data-sharing initiative.  Jost Zetsche, a German translator and localization and translation consultant who has written a number of books and articles on the technical aspects of translation, and is ATA’s leading guru of translation technology, moderated the discussion. 

Two large screens showing a robot-like arm clenched in combat with a very muscular human arm were set up at each end of the long stage.  To this commentator it seemed obvious from the beginning that we were about to witness a theatrical production of sorts.  The “discussion”, which took the form of a question and answer session with Zetsche asking the questions and members of the panel giving their opinions, appeared to have been scripted to make certain points that would lead to certain conclusions.   Here are some of those points and conclusions:

·         There is not just one market; there are many markets.  These markets offer translators the opportunity to earn more money.  Pre-editing (improving the source-language text) and post-editing (improving the target-language text) are rapidly expanding segments of the translation industry.

·         Data has replaced memory in importance.  Both MT and TM require a lot of data.  Translation “in the cloud” allows data to be searched efficiently and output is improved.  The hostile attitude toward machine translation has changed. 

·         Human translators need to communicate the value they add and, thus, there is a need to find a way to document quality.  Could that way to document quality be ATA “certification”? 

So, what was billed or advertised as a “battle” that should have involved genuine questioning, discussion, and differing opinions, turned out to be nothing more than a session clearly designed to enhance and promote the corporate image that ATA ceaselessly seeks to project:  “We are all united, we are of a singular mind and determination, and we are in agreement that the future is bright”.

In response to a recent article published here, one reader commented:  CAT tools themselves are not to be regarded as our enemies.  They didn’t do anything.  But the supposed translation specialists did it to themselves and the “real” translators let them.”  It seems to me that now is the time when we should all be questioning and discussing what is happening.  By way of one small example, translators are being asked to “translate documents that contain MT/TM and ‘new words’.  How should they price that kind of work?  Discussions on issues like this were totally absent in Denver and my guess is that they will be totally absent forever from the ATA scene.

Even among those who say they believe that translation is a writing profession, many have in reality forsaken that belief and adopted the view of the translation technocrats who foster the notion that translation is nothing more than the selling of “well-written” words.  Clearly, somewhere along the way the idea of translation as meaning and communication has been lost.  It certainly was lost at the 51st Annual Convention of the American Translators Association.