Translation memories (TM): the reality
Although these programs may be called translation memories, some of them are designed with the manager in mind more than the translator.
Since TMs allow the user to instantly find parts of previous translations, it can be tempting to use them to automatically translate parts of a text and just give the rest to a translator (which some translators call “translating the holes”). This might sound like a good idea, but a word of caution: if you don’t know how to use the tool properly, the consequences can be disastrous.
First of all, any text—whether a press release, annual report or user manual—is more than a patchwork of words or sentences. It is put together following a specific logic and structure, which are hardly random. The text forms a whole, and that’s what makes it effective.
Translating just a part of a document can ruin this unity and completely nullify the effort you put into giving it punch and pizzazz.
Second, some people think they can keep human translation to a minimum by putting anything they come across into a TM, regardless of quality. As a result, the larger the TM, the more polluted it becomes with translations that are either poor in quality or disconnected from the document to translate.
For a translation memory to be truly effective, it must contain only quality-assured translations, as well as be regularly reviewed so as to purge irrelevant sentences. Because this is a lengthy process, it is almost never done. Over time, translation memories become monstrous and out of control, sometimes yielding dubious results.
Finally, all this comes at a price. Technology can offer us powerful and effective tools… but only if we know how to use them properly. If not, instead of those hoped-for savings, you just might end up with a translation that needs to be rewritten from scratch.
(to be continued)
Translated by Joachim Lépine, C. Tr