Merge back 1 xlf to several target languages

27 views
Skip to first unread message

Víctor Parra García

unread,
Sep 14, 2024, 5:44:55 AM9/14/24
to okapi-users
Hi folks!
I have found a situation that I've not been able to handle through Rainbow/Tikal.

The context:
I have 1 whatever file I convert to xlf for translation either using Translation Kit Creation via Rainbow or just using Tikal.
As you know, for doing this we need to state both source and target locales so the xlf generated already have language codes.
Now I send that xlf file translated into different languages.

The question:
Is there anyway I can do the merging back to each target file for each language avoiding creating a Translation Kit for each target language to speed up both the xlf creation and the target merging back?

Thanks a lot!

Chase Tingley

unread,
Sep 14, 2024, 5:28:25 PM9/14/24
to Víctor Parra García, okapi-users
Hi Victor,

With tikal, I think the following works:
  • extract the XLIFF with one language pair (for example, -sl en -tl es).  (I would also run with -nocopy so it doesn't prepopulate the TU targets)
  • copy the XLIFF as many times as needed, update the target language metadata in the header and on the TU targets
  • merge each XLIFF with the source file to produce a target file in the desired language, using -sl en -tl ja (or whatever other language)
Tikal itself doesn't depend on any external state other than the source file and the XLIFF, and the Okapi merger only looks at the language codes in the XLIFF.  However, the language codes in the XLIFF do need to match the -tl you pass it on merge.

For Rainbow, I'm not sure how this works, I would need to double-check.  I believe the language code is also stored in the manifest.rkm metadata in the translation kit, and that might cause an issue.



Michel Farhi-chevillard

unread,
Sep 19, 2024, 5:32:54 PM9/19/24
to okapi-users
Depending on the CAT Tool or TMS you use to process the XLIFF, you may also be able to remove the target language codes altogether in the source XLIFF file and, with a little luck, your TMS will add that back with the proper language when it generates the target XLIFF.  I know that is what we do with Phrase TMS.   So, one common XLIFF for all the jobs into various languages, and several target XLIFFs in return with the proper language code.
Reply all
Reply to author
Forward
0 new messages