Our data is released under various Creative Commons licenses.More information
If you love this content, please consider a donation.
Menu
In order to get some sentences with interrelated context and audio, I plan to write a (linux shell) script that takes a movie and two subtitle files as argument. The idea is to link the sentences from the subtitle files and have sox (an audio editing program) cut out the audio from the movie, by looking at the time given in the subtitle files. All these information would be fed into a file that can be imported to anki.
Now, couldn't this be used for the tatoeba corpus as well (text only, not the audio due to obvious licensing problems)? I guess there's some kind of bulk import on the admin side. I quickly checked the first subtitle page I could come up with, opensubtitles.org, and found this on their homepage:
"As a name suggest OpenSubtitles.org is trying to be as open as possible. You can code application, script, utility or whatever you think is nice. First check which applications exists here [a link], then follow this [another link]."
What do you think? Or did someone have the idea already anyway?
subtitles rarely match at the sentence level, since they have to respect a format that is readable in the scene timing. It's interesting to see double-subtitling in countries such as Belgium (where foreign films are sometimes subtitled in both French and Dutch). I noticed many times that the subtitles didn't match. Sometimes they were not even close, sometimes whole sentences are just overlooked for lack of reading time...
So I doubt this is interesting on the sentence level...
Using http://www.universalsubtitles.org/en/database, it's still can worth it to have some kind of application that diplay on two column the sub and its translation and let a user quickly "validate" those who match a sentence level, could be a good comprromise ?
For open subtitles, script of movies are themselves protected by copyright, the same as for a book you have both the text itself and the way the editor has chosen to cut the paragraph, the font etc. which are also protected. So though it is "open" in the sense of "they ease reuse by other" I think they are on a grey line in term of copyright.
I need to add that universal subtitle script is what we use for the video in the main page of tatoeba for non-connected user.
A feature to post a sentence from a video with an open license (!) would be a great feature, especially when it could also add the sound (although i doubt the level of quality due to compression and ambient noise). Of course such tool should add the apropriate tags and comments for the correct attribution of the source.
>although i doubt the level of quality due to compression and ambient noise
Tu regardes trop de films d'action...
:) mais non, je voulais seulement qu'on fait attention à ça. Peut-être c'est drôle d'avoir des phrases comme: « Je vais te tuer ! [pouff] »
[fra] « Je vais te tuer ! [poum*] »
Comment dis-tu ? http://tatoeba.org/fre/sentences/show/1299152
Sorry for pulling this up again, but I just wanted to say to things: First, thanks for all your comments! There are really some points worth considering. Also I wasn't aware of universalsubtitles.org - I'll check that out sometimes. Second, I guess eventually I will try to realize the proposed script, but it may not happen during the next few weeks. If I can achieve something that may be of use for Tatoeba, I'll leave another comment. While I'm at it: Thanks to everyone contributing here, this has become my one and only source for learning material - it's awesome to have native speaker audio and sentences that are batch-processable :-)