Translation Guy Blog
Translation memory is the heart of everything we do.
For translators and their clients, it means faster delivery, better quality and more efficiency at a lower cost.
The traditional translation environment is not really translation-specific, forcing translators to perform non-value-added activities and to use multiple software programs during translation.
Translation memory (TM) software supports the work of translation by making previous translations easily available for re-use and by automating many translation-related tasks. Translation memory slices up source content into segments before translation, matching them to similar segments during translation, and then links source and target segments after translation.
TM systems facilitate the systematic re-use of previously created translations in a careful quality controlled environment. Translators use automatic terminology lookup to check terms as they go. The system makes sure that no sentences are skipped, and that formatting is preserved regardless of final output. Most importantly, translation memory work environments simplify linguistic review, making the editor’s job easier, so quality increases.
So TM has become essential to translation best practice. But the biggest reason translation buyers like translation memory is because of the cost savings. Reusable phrases mean big cost savings, every customer’s favorite phrasein translation.
But the potential for big savings can also mean big disappointments, too. A translation buyer promised big repetition savings by a translation sales guy like myself (I do it all the time) can find lower-than-expected repetition to be a real budget downer. Because of the source material, sometimes clients and translation memories just don’t see eye-to-eye on repeated phrases. A translation memory can entirely miss the similarities between two documents that jump out at the client.
This is because translation memory tools are very picky about accepting just any repeated phrases. Even a single comma or odd bit of code can drive an entire repeated sentence into the unique word bin of translation from scratch. Sentences that appeared quite similar on the printed page can be quite different in the memory of the translation machine.
In fact, there are specialists who work their magic on translation memory to fine-tune and repair the memory itself, or to manipulate content to be processed by machine translation for improved outcome. Careful pre-processing can dramatically improve repetition rates. We once got a 5% repetition rate up to 95%. That one special case required hundreds of hours of manipulation data before it was loaded into machine translation. The lower the potential repetition, the smaller the translation memory, and the less worthwhile such interventions become.
Most translation projects come out on the short side for translation memory, further reducing the opportunity for big savings. But even if translation memory savings are not so big for these little jobs, the benefits to quality are jumbo-sized.