Translation Memory Appendix I Wordfast Classic

The natural complement of a segmenter is translation memory. Every time a segment is translated, it is stored in the TM. Thus, a TM is a database of Translation Units (TU). A TU records source & target segments, date of creation, languages used, and the ID of the TU's creator. It also has a usage counter that records how many times a TU was re-used. The more a TU is re-used, the more it is valuable.

Translation memory, mostly on technical documents, can save a lot of time, because Wordfast will recognise segments that were already translated and propose them - you only have to check, validate and move on.

When Wordfast has delimited a segment, it will scan the TM, searching for an exact or approximate match to the source segment. If a match is found, the TU's target segment (the recorded translation) is proposed. Wordfast will display a number, ranging from 0 to 100, that rates the degree of similarity between the document's source segment and the TU's source segment. A 100% match is considered exact. A match under 100% but equal to or above the (user-definable) |fuzzy threshold is considered fuzzy; beneath that value, it is considered a no-match and will not be proposed.

If a translation is proposed, pressing Ctrl+Alt+M (Memory) will display the TU that was found during the TM's scan. In the case of a fuzzy match, differences between the document's source segment and the TU's source segment are highlighted. The |TM management section contains valuable supplementary information.

If Wordfast has found many matches, pressing Alt+Right/Left will display matches with lower/greater match value.

Back to Wordfast Classic User Manual