What the … is a Translation Memory?


We recently posted about CAT – or Computer Assisted Translation – software and why it’s important to use a tool that matches your requirements. Today we want to introduce you to one of the best features of CAT software – the Translation Memory.

CAT software has many benefits

Before the development of CAT software, translators essentially needed the memory of an elephant to work on large corporate projects. These kinds of projects could (and for translators without CAT software, definitely do) quickly become really complex as more texts are added.

Let’s say that each department writes most of its own content, but all employees across the organization have access to a common text pool. This type of text must be consistent across websites, flyers, presentations, image films, business reports, and many other sources. Clients also send translators extensive reference material for some jobs with the request that past translations get incorporated. And so begins a scrolling marathon through dozens of files with hundreds of pages.

To work around this problem and make sure a client’s corporate language stays consistent, most translators use Translation Memories (TMs). These databases store the source and associated target language texts in a structured manner. Each TM contains a bunch of information, such as individual words, phrases, sentences, and paragraphs.

The process

When a translator gets a text, they import it into their CAT software, which breaks it down into individual segments. By default, most tools will split segments at the end of a sentence, as this corresponds to the industry standard. These individual segments can then be translated and stored in the client’s Translation Memory. If a segment shows up in a future translation job, the translator has access to these entries from past translations. And as an added bonus, CAT software usually allows these segments to be automatically transferred (or “propagated”) into the target text. Even if the sentences are only similar and not 100% identical, the translator can still refer back to the previous phrasing. This makes it a breeze to maintain consistent corporate phrasing.

Aside from the consistency, there are also some other excellent benefits. For one, translators can complete projects much faster, because repetitive content doesn’t need to be re-translated. The translator just needs to proofread it. For another, TMs are fantastic for the client as it allows service providers like Wordcraft to apply consistent, clear discounts. We use percentage-based discount systems, or “matches“: the more similarities between stored TMs and new text, the less the translation will cost.

What are “matches”?

Perfect match (100%): The phrase the translator is working on is identical to one in the Translation Memory. This means that it has been translated before and is most likely still accurate.

Context match (110%): You may be wondering how a 110% match is even possible, and here’s your answer! This type of match means that not only does the current phrase match the TM perfectly, but the phrase before and after does as well (hence the name context match). It’s basically like a matching paragraph. It lets the translator check that the context still applies and move on quickly – and the client gets a hefty discount on these matches.

Fuzzy match: A fuzzy match means nothing more than something in the Translation Memory matches something in the new text. Think a word, number, or sentence fragment. A simple example is the match percentage between “the Tribble is brown” vs. “the Tribble is white“.

And when you really get in-depth, we can also use TMs for very advanced and cost-efficient workflows, such as the versioning of websites and various software applications, which we’ll cover in future posts. Gone are the days of copying and pasting content! Here’s to a future of translation processes full of fuzzy, matching Tribbles!

Thank you for joining us again on this journey through the World of Wordcraft, brought to you by the letters T and M. Check out last week’s post if you missed it, and we’ll see you next week for something even more futuristic!