Re: [g-a-devel] Speech synthesis and i18n

[Mario Lang <mlang delysid org>]

Samuel Thibault <samuel thibault ens-lyon org> writes:

> The short story is that I've noticed that some people on a french
> mailing list would write "ubentu" instead of "ubuntu", and when actually
> meeting them, I would hear them indeed say "ubentu" for the name of the
> Ubuntu distribution instead of "oubounetou", just because they never
> heard it correctly.

This is a very common problem in circles that are using speech synthesis
only for output.  I think I've already seen about 10 different ways
to spell "Linux" due to this.  But it is not limited to special words,
speech-synthesizerisms are all over the map.  I've heard people pronounce
first names consistently strangely, just because their speech synth
happens to pronounce the name in an unusual way.  The actual problem
at hand is that people do not know how the words are spelled and do not
check manually very often (since this breaks the flow of work if you
are using speech).

> That actually leads me to a maybe-not-so-niche problem: some words,
> although written the same throughout the world (GNU, Linux, Ubuntu, ...)
> are usually not spoken the same in each country.  Sometimes people will
> follow their native language pronunciation, sometimes they will use the
> english pronunciation, etc.  Speech synthesis, however, always use the
> native language pronunciation, which results to the story above.

> The Orca screen reader has a pronunciation dictionnary that can be used
> to fix that, but newcomers just don't already know the pronunciation :)
> I'm thus wondering whether it should be pre-fed with some data that the
> usual i18n translators would provide in the usual .po file.

How is this supposed to work in mixed language environments?
If I am being read an english text, I'd probably prefer the english
pronounciation of "Linux", while in a german context, I'd probably
prefer the german pronounciation.

> Of course, there should be a comment in the .po file explaining
> that, something like:
>
> #. Translators: this is the spoken word for Ubuntu, i.e. something that
> #. will be spoken the way Ubuntu would be pronounced in your language.
> msgid "Ubuntu"
> msgstr "Oubounetou"
>
> What do people think about that?

I think this is trying to solve a problem that exists
at a different level, namely that correct pronounciation
of words in speech synthesis is a non-trivial topic and
often very context dependant.

-- 
CYa,
  ⡍⠁⠗⠊⠕ | Debian Developer <URL:http://debian.org/>
  .''`. | Get my public key via finger mlang db debian org
 : :' : | 1024D/7FC1A0854909BCCDBE6C102DDFFC022A6B113E44
 `. `'
   `-      <URL:http://delysid.org/>  <URL:http://www.staff.tugraz.at/mlang/>