Start a new topic

If 2 original subs are merged, the translation of the second disappears - and former revisions of the translated subs are destroyed

Big thanks to esther434 who asked me if deleting an original sub would affect other languages: I didn't know, so I tried merging 2 of the English subs of The Lion King - Be Prepared - Tay Zonday  (see comparision of revisions #3 and #4 of the English subs) after having downloaded the French subs (Revision 1) for safety's sake.

In the French subs, this automatically created a revision 2, allegedly uploaded by "retired user", where the second sub disappeared - as well as my translation of the title and description of the video.
Moreover, the subs got completely deleted in revisions 0 and 1 (though the translation of the title and description are preserved as they were in these revisions)

Furthermore, rolling back the English subs to revision 3 (with the 2 separate subs) in revision 5  did not correspondingly restore the French subs as they were prior to the merging of the 2 English subs.

Therefore, when original subs have been /are being used in translations, could Amara developers please:
  • Add a big warning on the editing page of original subs that deleting them can severely mess up the translations?
  • Do something to preserve all the revisions of existing translations as they are, should one or more original subs be nonetheless deleted?

Meanwhile, fellow subtitlers, let's not delete original subs when they have been / are being translated.

And download the translated subs first as a precaution if you think you might forget and do some serious reshuffling. In my experiment, I only deleted ONE sub by merging 2, so it'll be easy to fix. But imagine if I had done several such mergings?


1 person has this problem

Hi Claude,

Thanks for bringing this up.

Yes, the data model design causes any changes made to the number or IDs of lines in the original transcript affect the dependent translations. Erasing lines in the original would destroy matching lines in the translations; uploads and rollbacks made to the transcript can fork or completely destroy the translations.
 
Your suggestions are good ones, however right now developers are focused on refactoring the data model, which should be a long term solution for this and other problems. 
 
Thanks for your reply, Jules, and thanks to the developers for working on a long time global solution. However, in the mean time, could at least a warning be added to the transcription interface?

Often, people are still working on the original subs when others start translating them - see the longer videos added to e.g. the Newshour team or for that matter, to your Global Voices team: if volunteers see their translations disappear because of a change in the original subs, that'll be rather discouraging.

Update: and there is a type of Amara video pages that is particularly at risk: those made from YT videos that already have original closed captions generated by a plain transcript sync'd by the YT software. Often, uploaders just leave them like that, or don't edit them satisfactorily: see many CC captioned videos of  this YT channel. The software-generated timing is accurate, but the splitting in subs is not humanly logical.

So when an Amara page is created from such a video and these subs are imported along as 100% complete, often some people start translating them, then someone else decides to re-split the original subtitles according to human logic.

That's a particularly at-risk case. But of course, the risk is also there for humanly generated subs: esther434's  question mentioned in the 1st post arose in the context of her translating some English subs I'd done for Watch the Full 2012 Vice Presidential Debate into Spanish: she also started to fix the English subs, and filled in some passages I didn't get and corrected others I had misunderstood, but she fortunately wondered about how deleting and resync'ing original subs would affect other languages before doing it: apart from her own Spanish subs, this would also have affected the work done on Chinese, German and Italian subs.
Esther wondered and thus made me check because she is a pro subtitler. But so many of us volunteers, like me,  aren't.
Hello,
This is very much a concern for me since I work on both English and Spanish.  Having to go back and forth between the languages, especially, in long projects is so time consuming and frustrating.  For that matter, if I make a change in English, does this mean that I have to let everyone one else in the other languages know so they can go and edit the work in their language? Wow, this seems to be a never ending situation.

I know that there are many things Amara/Univ. S. have on their plate.  But, please, when this problem is solved may we be informed?

Thank you.
esther434
Correction to my description of the revision 2 of the French subs allegedly uploaded by "retired user" in the 1st post of this thread: the second subtitle didn't actually disappear in it: it only got unsync'd. In fact, in the transcript generated by this revision 2, that sub appears at the top, as:
Not Synced [Scar:]                   
                    Pour être connectés, vous serez connectés
and in the Syncing stage of the editing widget, it shows grayed, but at the right place (see attached screenshot, with a red rectangle round it, both in the timeline and in the list). I could just sync it (I didn't so far, to leave things as described here).

So it's not quite as bad as I first thought. Still, it's quite a bother if that should happen before a translation is completed, because you can't have the translation widget (with the original text showing) anymore.

But actually, we don't need Amara to translate existing subs: Amara's greatest feature is the widget creating subs from the video, not the translation widget.  OK, Amara's translation widget is handy for collaboration, but not all that handy if translated subs are going to get unsync'd every time someone changes their splitting in the original ones.

So until this issue gets fixed in Amara, people could collaboratively translate with Piratepad, indicating the URL for each translation in a comment to the video page, for instance, then download the translation and upload it to the Amara page for the video - as "retired user" allegedly did in this case. Piratepad or a Google doc or a wiki page: Piratepad is just nice because you don't have to create an ID there.
Hello,

Indeed, both of you make excellent points, this is part of the reason why we're planning long term changes to our system, so that these issues are no longer blockers: languages will be able to get worked on side by side with different timing and seeing the changes: it will be LOVELY!

However, keep in mind that in any translation work, changes made to the original have to be made separately to all the translations if they are already in process or have been completed: it is a challenge that all collaborative translation work faces. 

We will certainly bring out the virtual marching band when we announce the changes in Amara, and just in case, updates and progress will be shared here in the forums! 


Login or Signup to post a comment