Translation Memories: better, faster, cheaper translations

Have you ever wondered how a translator can get through a 20,000-word user manual and make sure that 500 occurrences of Turn on the computer always translate into Encienda la computadora and not once in Prendé la compu? Meet Translation Memories, a translator’s—and your—biggest allies.

What are translation memories?

A translation memory is a database of previously translated segments with the help of a computer-assisted translation (CAT) software. 

When a source text gets imported into a CAT tool, the software breaks the content down into segments or strings. These strings are usually sentences, but can also be clauses, phrases, or titles. When the translator writes the translation of a source string, this source-target pair gets stored in the translation memory. 

If a similar or equal source string appears later in the text, the translation memory will prompt the translator with the previously saved string as a possibly reusable match. 

There are three types of translation memory matches: 

  • A context match, also known as a 101% or 102% match. This type of match means that the source string is equal to the one previously translated and that the strings before and after it are the same, too. There is a high chance that the translation can be reused.
  • A 100% match. This string is the same as the one already stored in the translation memory, but the preceding or the succeeding strings are different. The linguist can rely on its accuracy but should pay special attention to whether the new context still warrants the same translation as before.
  •  A fuzzy match means that the source string is similar to a previously translated one, but not the same.  For instance, there is a translation entry for The summers are hot in Tampa. A few pages down the road, The summers are hot in San Diego appears. The software recognizes the second string as a fuzzy match. Depending on the context, the translator might reuse some of the already translated text. The level of fuzziness can be anywhere from 99% to 0%. Fuzzy matches below 70% are generally useless, and the linguist will translate the string from scratch.

How do you benefit from translation memories?

Here are a few ways in which you benefit from translation memories:

  • Consistency

With translation memories, language experts can process a lengthy document without going willy-nilly with repetitive strings. The same string will always be translated the same way, ensuring a high level of consistency throughout the content.

  • Quality

All the stakeholders in a translation project (the client, the agency project manager, the linguist, and the reviewer) work on one single source of truth. This makes sure that everyone is aligned with brand guidelines, tone, and style.

  • Price

It’s standard practice that repetitions and high-percentage fuzzy matches are discounted. You might wonder why they don’t come free of charge. The reason is simple. Even a 100% match still needs to be proofread to make sure that the previously translated target language equivalent fits the new context. But usually, this revision comes at a fraction of the cost you pay for a new segment.

Do all types of content benefit from translation memories?

To put it shortly: no. User manuals, guidelines, product specs, and contracts are the most fitting group because they are repetitive. 

Creative writing like marketing collateral, hero copy, and social media contain little repetition. By consequence, they yield little to no optimization from translation memory software.

So the translation memory is an ever-growing translation tool that needs no maintenance?

If only that were true! In reality, as a translation memory grows, so does the need to keep them nice and tidy. Over time, you might have done changes to your content style guidelines. You might even have launched a new business line where you address a different audience. If your translation memories do not reflect these changes, the translator will face dilemmas when finding different matches for a given source segment. So instead of speeding up work, it will lead to wasting time trying to figure out which entry is the more relevant one. 

Translation memories help defy the Triple Obstacle Triangle. They make things cheaper, faster, and better—all at the same time. And the more they grow, the more value they bring.  But only if they are kept clean and updated. 

Not sure how? Give us a call, and we’ll tell you more.