Using freetranslation.mobi to translate .po files (Was: google translating gpl2+ licenced documentation...)
[Holger Levsen]
> Hi,
>
> debian-edu-doc is a gpl2+ document, which is translated into several
> languages. Now Petter had the idea to feed this into google translations,
> using http://freetranslation.mobi and committed the results back into the
> debian-edu-doc svn repository.
This is an interesting legal question, with many facets. Here is some
more background information.
The <URL: http://freetranslation.mobi/ > contain this information:
FreeTranslation.mobi
Simple. Fast. Mobile.
Enter a word or phrase:
[ ]
[langauge] to [language]
Tip: You can enter freetranslation.mobi/en-es/ (or any other
language pair) followed by the word or phrase to quickly pull the
translation (e.g. freetranslation.mobi/en-es/love).
Translations Powered by Google
A Yup.mobi Site
© FreeResources.mobi
Why .mobi?
I wrote a small perl script to process through a .po file and pass all
completely untranslated text fragments to this service and store the
resulting translation (if it succeeded) as a fuzzy translation in the
.po file. The translation then need to be reviewed by a human before
it is used to generate the documentation in question. I checked in
rough translations in debian-edu-doc after first running this new tool
for en->nb and manually checking a few of the new translations.
There are no terms of use that I have found available from the
freetranslation site, nor any clear proof that it is using Google
Translate. The "Translations Powered by Google" can be an indicator,
but it could also mean that the site uses Google cloud services for
all I know. If the site uses Google Translate, it is not possible
from that page to know what kind of agreement exist between the
freetranslation.mobi developers and Google. In any case, I believe
that I as a user of <URL: http://freetranslation.mobi/ > only have to
consider my relationship with that service and the information
presented by that service on their pages, and not what kind of
contracts exist with its internal subsystem contractors.
So as far I can tell, the Google Translation terms are irrelevant to
me for my use of <URL: http://freetranslation.mobi/ >, and only the
implicit access rights passed when I put text in the form and get text
back is relevant in this case.
It is still an interesting question to consider. It is not quite
clear who get the copyright if a computer translates. I believe that
I only gave <URL: http://freetranslation.mobi/ > the right to
translate the text I presented to it and that it gave me the rights to
use the text in return. The fact that the strings translated
(individually, but also overall) are only small quotes and fragments
of the entire document, I suspect fair use can also be claimed here,
if the small text fragments in question can be said to be
copyrightable in the first place.
I have found a few interesting documents related to this topic. No
idea if they present valid legal points, but interesting reads anyway.
Check out
* <URL: http://en.flossmanuals.net/open-translation-tools/ch057_machine-translation-and-copyright/ >
* <URL: http://cjlt.dal.ca/vol6_no3/gow.pdf >
* <URL: http://blog.fxtrans.com/2009/11/is-google-translate-accurate-enough-for.html >
Anyone know more relevant discussions? I would guess this interest
the translation community quite a bit. :)
Since I wrote the autotranslate tool, I have discovered the pology and
apertium free software tools also capable of translating text. In
this context it is also interesting to discuss who get the copyright
of translations done by such tools. :)
I'll follow this list for a while via Gmane, so no need to CC me.
--
Happy hacking
Petter Reinholdtsen
Reply to: