Translation memory (abbreviated as TM) is one of the biggest features available in computer aided translation software. But what is it? TM acts as a sort of database of text fragments – each piece of text that you translate is stored as a fragment, and therefore added to this database. The task of translation memory technology is to store, provide access to, and help analyze all these fragments.
Each fragment in the database is a sentence or phrase that you have translated. This information can be re-used in the future. So when the same source text is repeated, translation memory technology instantly inserts the translation that you previously made. The translator does not have to retype or retranslate, saving an enormous amount of time and lowering the expense of translation (since multiple instances of the same phrase are charged only once).
However, for the process to be fully automated, the repeats have to be fully identical. If there are small differences between the first and later instances of a phrase, the software will show the translation of the part that matches. But you will need to edit the translation of the new part that you have not translated previously, in order to make sure that it is correct. This semi-automated workflow ensures consistency among similarly worded sentences, speeding up and simplifying the process since the same phrases are used throughout.
Another important plus of the TM approach is consistent style and terminology. If a particular word has several potential translations but you prefer only one of them, translation memory remembers which version you select and will use it everywhere in the text.
That said, there are still certain limitations worth keeping in mind. Many words and terms are context-dependent. It may happen that particular words or terms simply do not fit a particular context. You should always double-check the results produced with translation memory in order to ensure that the result is consistently fluent, understandable, and accurate.
While the TM concept is common, different developers and versions use many different formats to store TM data. The most frequently used format is Translation Memory eXchange, abbreviated as .ТМХ. This format is universal because it uses the XML standard, which is popular and compatible with most TM software. So a TM saved in .TMX format can be used with many different applications. Even if you use one translation package at home and a different one at work, you can still exchange data and get work done using both programs.
However, not all translation software includes support for standard formats. Some developers favor proprietary formats that are not supported by others. Such applications include Wordfast and Deja Vu. This approach has its drawbacks but is encountered with a number of applications.
There are other XML-based formats as well. One of these is .ТВХ, which is currently undergoing the process to become an international standard. It standardizes data presentation and supports interchange with a minimum of information loss.
The slightly less widespread .SRX format is beginning to displace .ТМХ thanks to its power and high efficiency. Translation technology using this format tends to find segment matches better and more quickly.
The GMX GILT format is for complex large projects, primarily for companies that require top results even in the trickiest circumstances. Equipped with this format, software for computer aided translation becomes an incredibly powerful tool capable of meeting all conceivable needs.
Other, lesser-known formats exist as well: OLIF, TransWS, xml:tm, and XLIFF all offer intriguing features. However, they have not matured to the point of being able to seriously challenge the established market leaders.
So as you can see, translation memory technology is a rapidly changing field affected by the choices of CAT software developers, who are continuing their attempts to standardize formats. Learn more at
The SmartCAT translation platform supports all of the key industry formats, including SDLXLIFF, XLF, XLIFF, and TMX. To get started with this state-of-the-art localization tool, simply sign up, upload a file to translate as well as any existing TM that you have, and start your project.