Translation memories (TMs)
Translation memories (TMs) store translated sentences for translators to look up previous translations. The ability to reuse previous translations is particularly useful if a company has documents that are translated a number of times in different versions, for example if similar text is used in many documents. When, as a translator you translate some text, you can add the translation to the TM, so that later you or other translators can reuse that translation.
TMs and language pairs
The basis of a TM is to hold text translations from one language to another language - a language pair.
TMs in SDL GroupShare can contain multiple language pairs.
The TM divides text into segments and for each segment stores the source, the translation and supplementary data, such as date of translation. This collection of data that is stored for each segment is called a translation unit (TU).
A translator can use a TM to find translations for text. If a text segment to be translated is the same as the source text in a TU, the TM provides an exact match translation. If a text segment is similar, but not identical, to some source text, the TM provides a translation based on a fuzzy match.
Because a TM holds supplementary data for each translated segment, for example date, author and status, the translator can evaluate an offered translation and can choose between translations offered by different TMs.
Where the source segment has an exact match in the TM, the translation editor can use the TM's translation text for the target segment. As you work through the source content, translating segments, you can accept or override the matches that the editor offers.
Typically, after the translation has passed some review level, the project manager enables newly translated segments to be added to the TM. In this way, the TM is continually improved with use.
You can use more than one TM at the same time.
When you use the translation editor to translate text, the editor takes each source segment in turn and searches all the specified TMs for matches for the source text. You specify the order that the editor searches the TMs, and you can also grade each TM for reliability.