Translation memories: don’t you just love them?

21 Oct 2015 | Translation Industry


As we go further in the digital age, and we use Google and Wikipedia as a crutch to tell us who did what when because our memories and attention spans aren’t trained enough, specifically because we use Google and Wikipedia all the time, well- a translation memory is a good opportunity to expound philosophically on the idea of a perfect memory.

However, I won’t regale you with that 🙂 Instead, I will rave about translation memories.

What are translation memories? For a closer look check out this article. In short: they are a live segment (phrase, sentence) dictionary, and they are part of any good CAT tool. Here’s the best bit: they help translators spend less time translating, and they help you spend less money on translations.

The case of Kiabi

For our client Kiabi, we use our CAT tool MemoQ exclusively for the two reasons above.

Kiabi is an excellent example because e-commerce sites go along very well with translation memories.

As we receive a pretty high volume each week – 5 files per language per week for 3 languages (English, Spanish and Russian) – that’s 15 files per week with a volume between 500 and 4000 words per file, more or less. This requires several translators and a way to keep homogeneity between their different translations – this is where the translation memory comes in (alongside the glossary and term base).

We load the Excel files into MemoQ and it parses them, breaking up the product descriptions into handy segments. If translator A receives a translation including a segment previously translated by translator B, translator A will know she needs to keep the same translation, for consistency’s sake, and she will also not spend time translating all over again – this is how we save the client’s money, not to mention everyone’s time.

More specifically in the case of Kiabi, when product titles and part of the product descriptions are recurring (as is often the case), the client pays less for those.

Overall, 30-40% of the segments in one file will be repetitions or very good matches (i.e. previous translations that the software recognizes as almost perfect “dictionary” entries). They require minimal attention from the translator. The remaining 60-70% of content to translate is paid at full price.

This happens as the translation memory matures: you get a pretty good percentage of segments that you pay at a reduced price.

The translators then get more time to allocate to other projects, so it’s a win-win situation.

Obviously, there are situations where a translation memory will not be suitable to your needs, but they are applicable to many areas, so please ask for our advice and we’ll help you get excited about the possibilities.

Are you willing to give CAT tools and translation memories a try after reading this? Let us know in the comments!