TM management improvement #15316
TM management improvement #15316
Conversation
| origin=origin, | ||
| source_language_id=source_language_id, | ||
| target_language_id=target_language_id, | ||
| ).delete() # delete old entries, including those with different targets |
There was a problem hiding this comment.
This has risk of deleting unrelated entries as it matches only based on the source:
- There can be different entries from different components.
- There can be multiple same strings in one component, differing in key/context only.
There was a problem hiding this comment.
Doesn't filtering by origin exclude entries from different components ?
There was a problem hiding this comment.
As for the same string but different key/context, we can introduce a context variable to the Memory.
But going this route raises another question:
- Should we suggest a memory if it has a different key/context than the unit being translated ?
- If so, should we also apply another penalty factor for those ?
There was a problem hiding this comment.
You're right with origin, sorry for not realizing that.
Either context or a foreign key to Unit is needed to not delete entries for other strings. #9947 also describes a feature request to include context, so that is probably a better approach. Therefore, I've assigned it to you. Using the penalty approach for context mismatch is probably a good idea. I think 0.95 would work well in this case. We still want a 100% source match with context mismatch to rank above any machine translation.
gersona
force-pushed
the
7346_TM_management_improvement
branch
from
03317c8 to
0062486
Compare
Addtional question