AlexPGP's Corner

Ever since college, when it became clear to me that actually doing the homework was an important step toward getting an A or a B in a course like fluid mechanics, I've developed an informal set of tactics to turn stuff I'd rather not do into a game.

I do this now with some translations, and it generally stands me in good stead. One technique that I use is to track my progress through a document. The diagram below shows the number of words I had left to translate in a document against local time.


The gap between 10:40 am and 11:20 am occurred as I tried to deal with a small hurricane of tags in my segmented text.

To explain, Word files can contain tons of hidden information in text (this most often occurs when the file was created from a PDF). Since it's hidden, nobody generally gives a rat's tail about its existence. But when such a file is opened by a translation memory program such as memoQ, the result is pretty ugly:


All of those little gray doohickies represent some kind of instruction in the file text, along the lines of a change in font, or font size, or something along those lines. Translation memory programs that use such doohickies (the technical term for which is "tags") pretty much require them to appear in the translation (else the translated text runs a high risk of not looking right), and you'll pardon me if I don't bore you with the million and one ways satisfying this requirement can go wrong when there are this many tags in a segment.

How does one get rid of tags? Well, there are a number of methods out there, and none that I've found are perfect. The one I like the best is a set of Word macros marketed by a fellow named Dave Turner under the name CodeZapper (a copy of which was bought and paid for by yours truly some while ago). After running the basic tag-zapping macro, the text in the above illustration turned into this:


You'll notice there are a lot fewer tags in the cleaned up text, and while I could probably use this text as is, there were some other segments in the text that still retained a liberal quantity of tags. So I ran the heavy-duty zapping macro and got this:


Now, this is what I'm talking about!

The end result was mostly free of tags, and was a pleasure to translate.

P.S. For those impatient to know what kind of fascinatin' stuff it is I translate, here's the English translation:

The unit has a two-cylinder, four-stroke Briggs & Stratton engine, rated at 18 hp. The average fuel consumption (using unleaded gasoline) is 5.5 liters/hr.

Increasingly, agencies run new documents past their databases of document "segments" (typically, sentences) that have been translated before. Heck, I do this all the time myself, with databases of segments I've translated previously, as that is the entire point of the "translation memory" concept.

The problem arises when they embed found segments in documents sent out for translation and insist on paying a reduced rate for "editing" said translations.

Traditionally, editing a translation pays less than writing the translation to begin with. And that's because of the unspoken assumption that the translation has been competently written.

What is a competent translation?

Well, back when I worked in-house in Houston, and had a voice in defining "competence" in this context, we decided that a translation containing more than one major error per 250 words was poor, and one with more than two major errors per 250 words was unacceptably poor.

If one assumes that an "average" sentence consists of 25 words (an assumption that might be called "conservative," in the engineering sense), that works out to one major error per 10 sentences for a poor translation, and one major error per 5 sentences for an unacceptably poor translation.

So if you consider that a "pretranslated" segment that is less than an 100% match to the corresponding bit of source text is guaranteed to have a major error in it (after all, it's not a 100% match!), and that typically, less than 1 in 5 pretranslated segments are a 100% match, the end result is like editing an utterly and completely incompetent translation, for which one is offered a rate typically offered for editing a competent translation.

So in my current assignment, I'm not only finding the less-than-100% matches requiring attention, but the 100% matches as well. To wit:

{0>
Такая укрупненная частица (флок) осаждается более быстро.
<}100{>
Sedimentation of a large particle, or floc, is faster.
<0}
My version:
Such an enlarged particle (floc) settles more quickly.

{0>
Это происходит за счет совокупного действия нескольких процессов.
<}100{>
The effect can be explained by the following processes.
<0}
My version:
This is due to the combined action of several processes.

{0>
Параметризация турбулентности основана на сформулированном в (Озмидов, 1986) подходе.
<}100{>
The turbulent parameterization is based on the Ozmidov’s approach (1986).
<0}
My version:
Parameterization of turbulence is based on the approach formulated by Ozmidov (1986).

{0>
Турбулентность может быть представлена на основе вложенных вихревых структур разного масштаба, определяемого океаническими процессами.
<}100{>
The turbulence may be presented as the enclosed eddies of a different scale depending on oceanic processes.
<0}
My version:
Turbulence may be depicted on the basis of embedded vortex structures of various scale, determined by oceanic processes.

In addition to being annoying as hell, this kind of stuff slows me down significantly (I had, in fact, expected higher quality and factored it into my turnaround estimate).

Ah, well, there's nothing to be done about it now. I shall have to be more careful in the future, though, about accepting work from this client.

Cheers...

P.S. I repeat myself, I know, but it's therapeutic, okay?