One moment please...
You might have heard about translation memory, but do you know how it works?
Posted by

The nuts and bolts

A translation memory is a tool duomedia uses that enables the storing of translated content and its reuse. It takes the form of a bilingual database that contains a source sentence and its corresponding translation, and it is divided into segments. A segment typically consists of a sentence, but also can also be a whole paragraph, a title or an element in a list.

This memory is dynamic and it is continually updated with new translations. Concretely, this means that the more content you translate, the larger your translation repository becomes.

How is this different from machine translation?

With machine (or automated) translation, a computer will piece (concatenate) sentences together from previous texts, but the content is translated without reference to context and it is not checked by a professional translator. This process will give you the notoriously whimsical results of Google translate – would you trust an airline who tells you that they “take your bags and send them in all directions."? I wouldn’t either.

Translation memory in action

When translating a new document, duomedia starts by analyzing new content against the existing translation memory. This step enables us to determine the scope of work (how many words are there?). It also gives us a breakdown of the segments into three categories: new sentence, full match and “fuzzy match”.

Let’s take for example “the elephant in the room”.
If it has no match, it is a new sentence and will need to be translated from scratch to be added to the memory.

If “the elephant in the room” comes up in your text again, it will now be a full match in the memory and it won’t have to be translated again.

Now, has the “elephant in the room” morphed into a “cat in the room”? This cat will not be in your memory, will be identified by the translation tool as a fuzzy match and will need to be edited by a translator.

As a result, with each new translated document, new segments are stored in the database and will be available for reuse in future projects.

Why is translation memory good for you?

Using a translation memory can reduce turnaround times by reusing content that is already translated. The reuse of previous translation enables a consistent use of terminology and style across files and increases overall quality of the content. The content of the memory can also easily be exported and used by other translators.
Best of all, since fuzzy matches and full matches cost less than new sentences, your cost will be reduced.


Interested to know more? Contact us at

Message us