How Translation Memory Works

As we have mentioned before, the use of Translation Memory (TM) software could offer potential and significant savings depending on the nature of the text, the volume of the content needed for translation and the frequency of the translation needs.

While the above is true, it usually takes time to build a solid and robust TM because it differs greatly from a Machine Translation or Automated Translation Software which automatically translate a document without any human contribution, faster than the speed of light. We´ll clearly explain how Translation Memory works in the least technical way possible. Promise!

In a nutshell, Translation Memory is a database of all your translated text (source sentences and their corresponding translation). It is produced by translation software that you can buy from any reputable company in the market. This software is used by a human translator to translate given documents or text.

When using TM software, the source document is broken into segments. Segments are usually complete sentences, not individual words.

As the translator moves along with the translation, the software starts to store the translated units or segments and begins to build a memory. It is important to mention that the software does not usually store isolated or smaller units of text as they can be translated differently depending on the context.

Example: The software will store translation from English into Spanish for “The elections are around the corner.” But it will not store the translation of the word “elections” or the collocation “around the corner” individually.

Once your organization has translated any type of content more than once, or it has built a robust translation memory, the next time the translator translates new material the software will start to “suggest” which existing translated units or segments can be reused or be edited accordingly.

In other words, the software will identify previously translated segments and present the translator with matches which could be from 0% to 100%. At that moment, the translator proceeds to disregard, complete, improve, or accept the translation suggested. However, keep in mind that even 100% matches may be disregarded because everything depends on the context in which the sentence is being used.

A segment in the TM that is identical to a segment in the new text for translation is considered a 100% match. This means that at some point in the past, this exact segment was translated and stored in the TM. Usually, higher percent matches are closer in content to the new sentence for translation, and matches below 70% are not very useful for translators.

Those segments without a match (new segments) that are translated for the first time are automatically stored in the translation memory.

It is important to mention that in most cases translation memories become an asset for corporations.

Finally, keep in mind that the use of a Translation Memory on it is own does not guarantee the accuracy of the translation. The proper Translation Quality Assurance procedure needs to be in place. If you work with a professional and trusted translation provider, the use of a TM will make sure that translation is complete. It also will ensure that there is consistency in terms of terminology across the board in all your materials.

Language Concepts Consulting is a boutique provider of multilingual translation services. Languages we translate from and into include: English, French, Chinese, Korean, Portuguese, Spanish, Vietnamese and many more. We partner with organizations looking to penetrate and/or serve the multicultural markets in the USA, and Latin America. We provide native-quality and culturally-sensitive translations that get our client’s message across and resonate with their target audience.

For more information, please give us a call at 480.626.2926 or visit us at

Comments are closed.