Domain classification by project - still accurate in todays date with NMT integrations etc that are based on domains?
Old topic that always bugged me ....
We define a domain by project and that applies the domain to all files included in the project.
I believe that to be a fundamental error.
It also more prominently highlighted with the now more common use of domain specific NMT etc.
Currently it requires you to create subprojects by domain to increase the accuracy and efficiencies of assignment filters, TM Mix and NMT deployment.
It also causes your translation assets to be potentially wrongly tagged - hence retrain nmt based on wrong domain.
Domain identification should at least be identifiable down to file level, if not even down to segment/paragraph level to be effective for daily use.
Step 1 could be moving domain identification or adjustment option down to a doc level (add field Domain), allowing the PM to adjust the Domain identification on a document level - hence assignment policies etc would be more efficient/accurate
Step 2 Sliced files
Allow slicing files (non destructive) and enable domain identification on slices of documents - even more accuate
Feedback/ Suggestions ?
Customer support service by UserEcho