DTP and Translation Memory

A couple weeks ago, Bill Kelts in Production wrote an introduction to Translation Memory (TM). With this post, I'd like to get into how TM affects your project on the publishing side.

Let's say we translated your sporting goods catalog last year. Here is how your top-selling product looked back then:


This year, however, you revamped the layout a bit and wound up with this:


Shouldn't matter, right? It's the same text, and it's already been translated, so we'll have 100% consistency, right? Not so fast!

How did your artist execute this runaround effect? Let's start by looking at last year's version, with all the layout marks visible in InDesign:


And this year?


Maybe the basketball graphic didn't have a proper clipping path embedded; maybe InDesign's "wrap to object" feature wasn't accurate enough; or maybe the person who did the layout doesn't know what either of those two terms even mean. Whatever the case, a TM tool does not perceive these as being the same sentences. In fact, they are so different the text in them won't even show up as partially leveragable "fuzzy matches," but as completely new text...potentially to be charged to you at the unique word rate.

Here's how the original and the altered text blocks look to a translation tool:


We went from one complete paragraph to 12 segments - with no text being recognized as repetitive or a close (fuzzy) match to text already in the TM. Repeat this problem throughout a 24-page catalog, and you've lost a lot of time and money. There is also a high risk of inconsistency between last year's catalog and this year's, which increases QA and revision time.

We in LinguaLinx's DTP department do everything we can to clean up files before they are run through the text-extract and translation process, but this takes time and adds production hours (and dollars) to the project.

In subsequent posts, I will address additional specific examples, as well as general strategies for creating localization-friendly layouts in InDesign and QuarkXpress.

What is TM and why do I need one?

Translation memory (TM) is exactly what you think it is, in the simplest terms: A database that stores previously translated segments. This translation memory aids professional linguists in keeping production time low and consistency high.

The key word in this definition is segments. When the term "translation memory" is mentioned, often people get confused and think that every translated word is stored individually inside this memory. However, this is not the case. A TM will consist of translated segments, sentences, and paragraphs. Many people feel that translation memories were created for the sole purpose of reducing costs. This is also false. Although the use of TM can cause a slight decrease in price over time, translation memory was created to help maintain consistency and quality throughout documents. TM ensures that mission statements, slogans, and names will be translated consistently throughout all documents for a particular company.

When translating for the first time, do not be too concerned with the potential price benefit of translation memory. Instead, consider for which translations TM is best suited. Over time you will see the benefits. Remember that price alone should not dictate which translation provider you choose. Consider the quality translation that may be gained for virtually the same cost in comparison to not using TM. Translation memory is here to help you keep brand consistency. And that consistency is invaluable to your business.