problem

when using t5memory TM matches with tags, it may happen, that the tags send in the source could not be applied to the target answered by t5memory. That is depending on the content in the TM, and is done correct so far by t5memory.

It is also correct that t5memory returns such segments as a 100% match, since the matchrate reflects the comparsion of source content, which is equal in this case.

Finally that leads to segments, pre-translated as 100% match, where tags are missing in the target, or having an invalid tag structure in the target.

solution

Setup of test data:

• add the attached TMX files as separate language resources.
• Case2 file triggers the here described issue.
• Case1 would be another situation to be discussed if we can solve that somehow on our side or not.

After re-applying the internal tags to the t5memory result and removing additional invalid tags, we have to apply the tag cleanup to the segment content, so that we save no invalid tag structure into target (and target at import time). Missing tags are revealed by the QA check, but invalid tag structure (especially in the target at import time field) may lead to subsequent errors. Here it must be ensured that the tag is invoked also on locked pre-translated tags.

Attention: This should only be done for automatic pre-translation. When taking over matches in the UI, this is done implicitly by the UI. In the match panel the match is shown with the additional tags / wrong tag structure. When taking it over, the additional tags are removed, and an invalid tag structure is in the editor which must be repaired by the user.

Still that opens a hole wenn a translator is allowed to save tag errors and saves then an invalid tag structure which is then placed in target at import time again. Or is that not possible anyway?