What Is Translation Memory, and How Does It Work?

At Mission Translate, we often talk about the importance of working with professional human translators to obtain high-quality results for your translations. However, that does not mean that we do not leverage technology to maximise the efficiency and consistency of our outputs without jeopardising their quality.

There is a crucial distinction between machine translation tools, where human input is minimal to none and computer-assisted translation (CAT) tools, which translators use as they work to help them achieve the best quality result. Embedded within most CAT tools is a function called Translation Memory, an excellent aid, particularly when working with large volumes of text, where there may be repetitive content.

But what exactly is it, and how does it work?

What is translation memory?

Translation memory or TM is the digital memory within a CAT tool that stores source segments, such as sentences, paragraphs or sections of text, which have been matched to their translated partner by the TM tool. Each stored segment in the translation memory includes the original, source language and its translation or target language.

Storing these language pairs, called translation units, and making them available to the translator while working enables them to work more efficiently and maintain consistency within and across their translation projects.

What are the main features of translation memory?

Two key processes carried out by translation memory are:

• Identifying linguistic segments from a source text that have been previously translated
• Memorising and categorising segments, marking them according to their match type

How does translation memory categorise the different types of matches?

Translation memory recognises that content can be closely matched but not always exactly the same. Therefore, when reviewing a text for translation, it identifies segments and categorises them depending on how closely they match the stored memory. 

Exact match

An exact match (also known as a 100% match) is when the current source segment matches character by character with a stored one, indicating it has been translated previously.

Nearly exact match

With a nearly exact match, the source text of the segment is almost the same as the match (95-99%), but there are slight differences, e.g., numbers, tags, punctuation marks and spaces.

Fuzzy match

A fuzzy match is when a partial match is identified between the current source segment and a stored one. Sometimes percentages – greater than 0% and less than 100% – are attributed to the extent of the fuzzy match. For example, high fuzzy (85-95%), medium fuzzy (75-84%) and low fuzzy (50-74%).

Exact match with context or in-context exact match

With this match, the source text in the segment is exactly the same as the match from the translation memory. Additionally, the context of the source text is also the same as the context stored in the translation memory. It is, therefore, sometimes referred to as a 110% match.

How does translation memory improve the translation process?

By identifying and marking the categories of matches within a text, translation memory speeds up the translation process, saving the translator time in researching and manually translating the segment. It also improves consistency, enabling the translator to use previously translated segments as they work through the document.

The use of translation memory is not exclusive to one file. It builds its memory over time and across multiple documents. As its memory grows, the more segments are matched, and the more extensive the benefits become.

What quality control processes are in place?

At Mission Translate, all translators still check all translation segments and confirm that the matches are accurate. Even when matches are 100%-110%, we require our translators to check them manually and set the colour to green, indicating that the segment has been verified. If the segment is not verified, the translation cannot proceed to the Q&A checks, which include proofreading from an independent translator and internal confirmation.

What does this mean for a client?

The improved consistency of a translation means a higher-quality document for the client. Each client will have their own translation memory, storing translated units for things such as technical terms or a brand’s tone and style. These can then be accessed any time a new translation is required.

Using translation memory to improve the speed of a translation also leads to a reduction in cost. These discounts are typically passed onto the client through the percentage of matches found.

At Mission Translate, we always seek to increase the size of our clients’ translation memory, enabling them to receive the best possible value for their translation project. When providing a quote, we are transparent in our breakdown, indicating the different types of matches and the discounts offered.

To get a quote for your next translation project or discuss your requirements, please contact our team at [email protected].