Title: Very high match rates when the segment to translate is very different from the source segment in the translation memory
Description: In some cases, memoQ can return a TM entry as a high percentage match for a source segment that is very different. Consider the situation on the below screenshot
Segment 2 was confirmed and saved to the TM before, and contains a number and a pair of tags. Segment 1 contains something non-textual, in this case a series of asterisk characters. The TM entry is listed as a 99% hit which is obviously not desirable, as there is no similarity at all.
The occurs under the following circumstances (all of the below criteria must be met):
- The translation source segment is Korean, Japanese, or Chinese.
- The TM was created in a memoQ or memoQ server version older than memoQ 2015 build 100 (7.8.100), and has not been repaired since then.
- The memoQ or memoQ server version being used is memoQ 2015 build 100 (7.8.100) or newer.
To resolve the problem, repair the translation memory
in memoQ (server) 2015 build 100 (7.8.100) or newer
. After repairing the TM
, the very different TM
entries are no longer listed as suggestions. (In the case of a local TM, the memoQ client version matters, in the case of an online TM, the memoQ server version matters.)