Translation Technology

[Jim Gettys]

The other *key* point to enabling translations is to have good
terminology defined in advance for a new language; otherwise, since much
of the terminology (or even basic vocabulary) won't exist in many of the
languages, we'll get back translations that have no consistency
whatsoever (in languages for which computers are new).  And this
terminology may need to exist in (a) language(s) other than English.
Example: in Peru they have delayed translating to Quechua or Aymara in
large part that most of the people who might do the translation speak
Spanish, rather than English.
                            - Jim

On Fri, 2008-06-27 at 17:46 -0400, Michael Stone wrote:
> Yesterday, Sayamindu, Jim, Chris, Eben, Dennis, and I met to discuss
> what technology we could provide to improve the experience of producing
> and consuming translations and other localization data on the XO.
> 
> We sought to address three questions:
> 
>  1) How can we make "some" (then "every") string translatable. 
> 
>  2) How do we distribute translations?
> 
>  3) How can we separate translations from builds so that new translation
>     improvements can be incorporated long after an operating system is
>     released?
> 
> Our conclusions were that we could satisfactorily address these
> questions by providing:
> 
>  * a disconnected on-XO translation interface.
> 
>      - Could be implemented as an activity named Translate. That
>        activity might:
>        
>          + spider the build for translatable strings in activities and
>            Sugar.
> 
>          + suggest appropriate terminology as possible.
> 
>  * for medium-privilege modification of translations.
> 
>  * for horizontal distribution of translations via some low-latency
>    network.
> 
>  * for vertical distribution of translations via some delay-tolerant
>    network.
> 
>  * a 'big undo' button similar to the olpc-update 'undo' mechanism which
>    permits experimentation with translations.
> 
>  * for customization of translations via our existing USB customization
>    technology and in the follow-on image-builder technology.
> 
>      - we discussed two implementation ideas:
> 
>          + to teach the customization infrastructure about translations
> 
>          + to teach the customization infrastructure to install RPMs
> 
>        but no consensus was reached.
> 
> We believe that supplying this technology and infrastructure would:
> 
>   * go a long way toward our long-term dream of permitting every child
>     using our software to learn in his or her native language(s).
> 
>   * greatly improve the ability of bilingual individuals using our
>     technology to autonomously improve the quality and coverage of our
>     translations regardless of their access to internet connectivity.
> 
> Michael
> _______________________________________________
> Devel mailing list
> Devel at lists.laptop.org
> http://lists.laptop.org/listinfo/devel
-- 
Jim Gettys <jg at laptop.org>
One Laptop Per Child