What is Translation Memory (TM)

Published on Oct 17, 2019

by Marcos Dinnerstein

Summary: Translation Memory (TM) saves modern translation platforms time and money by reusing material previously translated.

In the days before we had computers to assist us, translation work was a one-off, hand-crafted affair. (Bespoke, if you want to use the latest jargon) A translator did their work and that work had no relation to any work done after that, even when covering similar topics. But with the development of computers and specialized databases, Translation Memory Systems were developed.

Wikipedia defines Translation Memory as:
...a database that stores "segments", which can be sentences, paragraphs or sentence-like units (headings, titles or elements in a list) that have previously been translated, in order to aid human translators. The translation memory stores the source text and its corresponding translation in language pairs called “translation units”. Individual words are handled by terminology bases and are not within the domain of TM.

How do people use Translation Memory Systems?

The software segments the source text, i.e. the original text, into components and searches the Translation Memory for matches between segments in the source text and segments that have previously been translated. These ‘source-target’ pairs are what is stored in the database. When a match is found it is presented to the translator as a translation candidate. The translator can then, accept it, replace it with a new translation, or change it to match the source. In the last two options, the altered or new translation goes into the database.

This process lets translators easily reuse prior work done by anyone contributing to the database, that is, the Translation Memory. As a result, you have greater consistency, speed, and savings.

It won’t surprise you to learn that governments generate a great deal of translation work. The NEC TM project uses government-funded translation work to supply a database open to participating European countries. Why have these Europen countries joined forces on this project? Each is pooling their national bilingual assets as open-source data to be used to train machine learning. This allows the member countries to access a larger data set and make their spending on translations to be more effective.

How MotaWord’s Translation Memory Helps Our Clients

When a MotaWord client uploads a text to be translated, segments of that text are stored in MotaWord’s Translation Memory. Over time, as more and more translations contribute to the Translation Memory, translations become more efficient, consistent and cheaper because we don’t charge for any text that was previously translated.

  • Clients can download their translation data.

  • MotaWord does not charge for creating a Translation Memory for clients.

  • MotaWord does not share a client’s Translation Memory

MotaWord’s clients appreciate that we don’t charge them for duplicate translations and can often turn around jobs more quickly.

Learn more

Wikipedia: Translation Memory

Why the EU and Companies Are Using Translation Memory in Boosting Productivity

About the NEC TM project

SUBSCRIBE TO OUR NEWSLETTER