This week is turning out to be an interesting one. Practically all work that I’m doing is reminding me of the negative consequences of the practices that are being adopted by some of the larger international translation agencies.
So far, I’ve worked on five different jobs this week. One of those was for a translation agency in Germany that is not, as far as I know, in any way linked to the bigger international agencies. The job went smoothly, no problems at all.
For the four remaining jobs I worked either directly with one of the leading international agencies or with one of their subcontractors. One of the jobs was fairly standard and presented no issues. The other three were a different matter alltogether…
Job 1 was a user manual for a sperm selection device used in ICSI procedures. A previous version of the manual had been translated earlier, but the translation had not been stored in the Translation Memory that I was supposed to use, because the end client had expressed concerns about some errors in the previous translation. My client (the subcontractor) did send a PDF version of that translation for reference, though, but later instructed me not to look at that too much, because the quality was ‘not very good’.
I started translating and discovered that the new version was practically the same as the old one. So, hoping to find some useful terminology and possibly some reusable translations, I did turn to the PDF anyway. I soon discovered that ‘not very good’ had been an understatement. I was absolutely shocked by the (lack of) quality of the translation. The text had obviously been machine-translated but not post-reviewed and the translation was completely unintelligible. When I mentioned this to my client and expressed the hope that this translation hadn’t been published, she informed me that the Finnish translator had made the exact same remark. Someone somewhere in the workflow had probably wanted to save some money by leaving out the post-editing round. Had they not done that, they would have been able to leverage the previous translation, making the new translation a lot cheaper.
Job 2 was a review job, directly for this large agency. The text was a speech by the CEO of an international Group, active in chemistry. The company is based in Belgium and the speech, in French, was to discuss the results of the financial crisis, an acquisition and strategic plans for the future. The translation, into Dutch, had obviously been performed by a Flemish translator, and not a very good one. The main positive thing that I can say about it, is that the translator must have run a spell check, because typos were the only type of error that I didn’t find. I had to make stylistic and grammatical changes to practically every sentence, and in several cases the translator hadn’t understood the French text at all. He or she had used incorrect verb tenses and had used literal translations for French idioms. The result was a text that could certainly not be used by a CEO as a shareholder speech. Fortunately, my client agreed to double my review rate to compensate for the additional time I’d have to spend to create a better translation. My guess is that their budget would have allowed for that anyway, as they’d probably used a translator charging rates that are far below the generally accepted minimum rates. If they’d had the text translated by a quality translator charging normal rates, a short spotcheck QA would have been enough.
Job 3 was for the subcontractor again. This time I was asked to translate software for a CPR defibrillator – the text that is displayed on-screen. Now, there were several problems with this job. First of all, the software had been extracted into an Excel file. Additional columns had been added with information about the meaning of the software strings and about length restrictions. A macro had been added to the file to check if the translated strings would comply with the length restrictions. So far so good. Unfortunately, the file had been sorted alphabetically, so any context information had been lost. Also, because of the length restrictions, many of the English strings had been abbreviated and it wasn’t always clear what the unabbreviated string would have been. There was a PDF manual with a little more context information. The file consisted of more than 1,500 strings, though, so I would have been required to look up 1,500 strings in the PDF to make sure I had translated them correctly.
That would have been bad enough as it is. But to make matters worse, the end client has their own online translation memory software. They require their translators to use that software for all jobs, one of the reasons being that in that way they are building up material to use for their machine translation system.
And so the column with translatable strings from the Excel file had been copied into a Word file that I would have to translate using that online software. All additional context information and all length restrictions had been stripped from the file.
Providing a decent translation would therefore take a lot of additional time. For each of the 1,500+ strings in the Word file, I would have to consult first the Excel file for some context information and then the PDF for more information. I would then have to think of a proper Dutch translation, and then have to consult the Excel file again to see if the translation fitted within the length restriction. If not, I’d have to abbreviate it without loosing its meaning. And all of this for the normal word rate, which isn’t all that high in the first place.
I did start on the translation. I’d taken it on without checking it properly beforehand, so I felt obliged the make the best of it. But the more I translated, the more I started to doubt. I felt that I could in no way guarantee that my translations were correct, or that they were still intelligible after I’d abbreviated them. If this had been some innocent piece of software, that would have been bad enough. But this wasn’t just any software, it was CPR defibrillator software, used in emergency situations where there’s no time to consult the manual. With incorrect or unintelligible on-screen instructions, people could die. I did not want to be responsible for that and returned the job to my client, who said they would find ‘another solution’ and mention my concerns to their client, the large agency. It’s my guess that ‘another solution’ means finding another translator who doesn’t have my concerns.
There are several things here that cause my concern. Obviously, the main problem is that this agency, as many of the large multilingual and multinational agencies, is trying to offer the lowest possible rates to their end clients. And because they still need to make a profit, they are using several methods to cut costs. They are using machine translation, as in job 1, but don’t always want to pay for a post-editing round. They are lowering their word rates to the extent that quality translators no longer want to take on their jobs, as was probably the case in job 2. And they are shoving all additional work on the shoulders of freelance translators without paying for the additional time needed to produce a quality translation, as in job 3.
In the first two examples, the end clients were lucky that someone noticed the quality issues and that something could be done. This led to additional costs for the agency, which they may or may not have been compensated for by the end clients. However, in job 1, a horrible translation had been produced and (probably) published first. And in cases such as job 2, it’s all too common for the review round to be skipped for whatever reason. The blame is usually put on translators in general. Asked what we do for a living, we have all been presented with examples of badly translated user manuals for watches, cameras, etc.
But job 3 shows that it’s not just the translator’s reputation that’s at stake. Bad translations could put people’s lives at risk. It is completely unacceptable that people could die because companies and agencies are trying to save money in the translation process, at whatever cost…