Start a new topic

When (not) to use Bing automatic translation

Automated translation is very useful to get a rough idea of what a text is about when you don't know the original language well enough or at all. And as Closed Caption (CC) subtitles are a text, they can likewise be automatically translated.

That's why the Amara translation tool also offers the possibility to automatically translate subtitles with Bing. However, as  Marisa Jean Browne wrote in the How do I translate subtitles? FAQ:

"Users suggest that this option only be used by those who are willing to correct mistakes and inaccuracies before saving and exiting subtitles, since it is considered bad etiquette to leave a machine translated subtitle for others to fix."

That's a bit tough, though, because there are heaps of things to be corrected in any Bing translation, so you might well have to save and exit before you've finished.  Workaround: explain what you're doing in a first "mock" subtitle you'll delete when you've completed the corrections.

More importantly: never use Bing to complete a subtitle translation that others are already working on normally, i.e. humanly: it is a royal pain to have to go through all subtitles to identify and correct the Bing ones. Vice-versa, if you are a human translator whose subtitles have been nevertheless Bing'ed by someone else, you can always roll back to the last purely human revision and carry on from there.

Then be aware that Bing cannot cope with long subtitle sets: see  ehazlett's second post in

Finally, if you want to get a rough idea of what a video whose language you don't master is about, without messing up in-progress subtitles, and/or if the subtitle set is too long for Bing:

  1. Download the original subtitles as .srt
  2. Have them automatically translated by the Google Translator Toolkit
  3. Download the autotranslated subs as .srt
  4. Upload them to the Amara video, but as if they were a translation into one of the non-existing languages of the Amara language list: Metadata: Geo, Metadata: Twitter or Metadata: Wikipedia. This way they won't interfere with real translation work.

Login or Signup to post a comment