Last week, I found myself suddenly rolling out a blog post because of an unpleasant experience involving machine translation (or should I say machine pseudo-translation), which I wanted to share with you all. This week I’m shaking my head at yet another affront to our profession, which sadly seems to have become rather commonplace judging by email job requests I receive and forum posts I read.
I’m talking about CAT tools, the discounts we’re expected to give as a result of using them, and the silly deadlines imposed in the belief they will magically improve our performance to near super-human levels.
Don’t get me wrong, I’m actually a huge fan of my CAT and I do absolutely every translation with it. Sometimes I even use it to do revisions (i.e. checking other translators’ translations against source, which is not proofreading, but I’ll blog about that later) so I can keep a record of interesting turns of phrase and useful terminology. Unfortunately, I resisted using CATs for a long time. I was one of those vociferous opposers who insist that segmenting a translation will spoil the flow and be detrimental to quality. That was before a reviser of one of my translations pointed out (to my huge embarrassment) that I had failed to translate repetitive text later in a document exactly the same as I had done at first. Realising this would never have happened if I had used a CAT, I bought my licence and I have never looked back since. However, I do feel that I missed out on the CAT boom in some ways. By the time I was using one on a regular basis, most agencies had cottoned on to these tools and started not only to use them, but also abuse them.
Fuzzy-match discounts must be a dirty term in our industry by now, especially if it’s somebody else’s memory inputs you’re expected to use. Any translator worth their salt is surely going to find it difficult to let an error-ridden segment pass them by without correction, regardless of how much (or how little) they are being paid for matches. Yet if we do improve segments here and there in a translation memory for a pittance, we are just further driving down prices and undervaluing our expertise.
Yesterday I was offered a 44,000-word job. Usually I would jump at the chance to land such a large translation, but the agency wanted it back in a day and a half. This laughably short turnaround time was apparently justified by the fact that only 7,000 words were ‘new’ (i.e. under 75% matches), and that the rest could be processed using the end client’s memory. Given that the subject matter was of interest, I decided to investigate the quality of the memory before rejecting the job outright, despite the nigh-impossible-under-any-circumstances deadline. And therein lies the rub. I analysed a fair few of the segments and decided that I didn’t agree with some of the translations of fundamental terminology. It was going to be impossible for me to overlook any 100% matches containing these (IMHO) mistranslated terms, so I refused the job making my reasons for doing so abundantly clear.
Some poor sod is slaving away at this translation as I write. Maybe they will read this post when they surface from delivering this nightmare job and know better next time. Meanwhile, I hope I have managed to remind and even convince a few of my readers that ‘no’ is a very useful yet often underused word that should nevertheless feature a great deal more in a translator’s negotiating vocabulary.
This post was first published on 20/11/2013 on my previous blog.