problem

We assume that language resources behave correct when dealing with tags. But this may not be the case.

Currently we have a problem (TS-647) where NEC-TM was used for pre-translation of a docx document. On pre-translation a x tag was added to a segment where in the source no X tag was at all. Since on retransformation of the xliff tags to translate5 tags no suitable translate5 tag could be found for the x tag, the x tag remained originally in the target. This X tag was invisible in the GUI.

On export the X tag remained in the xliff send to Okapi, which then produced an error since the document could not be exported with this additional tag.

Additional similar problem:

Also we had a case where a starting tag of a tag pair was missing from the deepL pretranslation. In that case we can not just add the tag add the end, since the tag order would then be wrong. How should we solve that? Just log a tag missing warning? Wait for AutoQA tag check?

Additional problem, that could be related

See TS-699

reproduction

• Generate a docx document containing one sentence.
• Translate this document with pre-translation and inject additional x tags in the connector part of a Language Resource of your choice (for example the ZDemoMT in the same named plugin)
• ensure that the additional x tag is in the target then (by using F12) in the GUI
• try to export this document, this should trigger the Okapi error then

solution

• Additional tags, not able to be remapped to translate5 tags should be removed from the content
• Search Jira for similar issues and solve / set them accordingly!