Post date: May 26, 2011 11:55:39 AM
Carrying out a quality check on a batch of technical translations recently, I was struck by how easy it is for translation errors to be recycled via translation memory. I kept seeing the same incorrect translations reproduced over and over again, by different translators, in documents from different batches. This is not the fault of the TM software, but of the translators (and translation purchasers) using it.
It is as if a translation somehow becomes validated once it has been entered into translation memory. This is, of course, absurd: a bad translation remains a bad translation, irrespective of whether it has become infinitely reproducible thanks to the power of technology. It seems that no effective quality check is performed at a sufficiently early stage to prevent bad translation from being "set in stone" in translation memory.
Fundamentally, of course, translators should be capable of producing good work in the first place; however, they also need to be able to assess the quality of translated text (their own as well as other people's), and make appropriate corrections (if necessary in liaison with the customer). The notion that existing translation memory cannot be "touched" by translators needs to be challenged.
Think about it - when you import the translation segment into your file, you can see both the source language and target language versions, so you should be able to tell whether it's rubbish or not. Letting rubbish pass unchallenged is as bad as producing rubbish in the first place.
This is potentially a serious issue for translators and translation users.