| |
| |
|
 |
This enclycopedia is a work in progress. Please do not hesitate to email me if you disagree with this article or found an any error in this page. |
 |
 |
 |
 |
|
|
translation memory
[Article still incomplete]
Set of text segments associated with their translation. The segments used usually are paragraphs or sentences. The idea is that during or after the translation we keep in memory of how each segment was translated so if the same segment is encountered in the future it does not need to be translated again.
While this technique is very efficient on text that have been updated, it is a bit less effective on new text. Indeed aside from standard expression, it is pretty rare to encounter twice the same sentence/paragraph in different text (you can do the test in google: pick a simple sentence of a few words in an article and then search for it on the web).
It is easy to understand that the smaller the segments are the more chance we have of finding exact matches later. Of course if the segments are too small (e.g. a single word), there is a increased chance that based on the context the translation may become inaccurate (for example the wrong gender is used). Aside for wording differences, punctuation and formating differences may also degrade the performance of a translation memory.
When no exact match can be found the translation memory can be used to find the translation of segments that are almost similar. We then talk about fuzzy match. This may help the translator to provide a more consistent translation.
|
Article submitted by: Thierry Sourbier
Last modified: 2008-10-07
Rate this article:
|
directory 
encyclopedia Beta