Re: [orca-list] Orca: Support for Multilingual text


I'm also curious about the configuration issue mentioned in some of
the relate bugs.  If I was implementing this I think what I'd do is
something like this.

Irst lets declare a struct for what options a voice profile has like

struct voice_profile{
char *name; // why not mane it
char *lang;
char *synth;
int rate;
int pitch;
... // some more stuff we'd like to have

Then we can keep a circular linked list of these structs and have
keystrokes to move around the circle.  We could use some other data
structure, but who likes corner cases.

Then we have to read and write this to a config file.  It sems like
this is fairly easy if we import the config fle like we do know, or
read and parse it, but I'm not sure I see an easy way to do it with
gconf.  I admit to not knowing much of anything about gconf or
gsettings, but it seems like it can only have mappings from a given
set  of keys to values?  For importing a file I'm assuming that python
allows one to statically declare an array of structs or classes like
in c something like this.

struct profile_list[] = {
{"main", "en_us", "espeak", 4, 5,},
{"alternate", "en_gb", "flite", 4, 5,},


On Mon, Jul 19, 2010 at 01:07:58PM -0400, Trevor Saunders wrote:

 As has been discused previously I think generalized voice profiles
are better than language profiles, the only real difference being that
you could have different profiles for the same language.

As for auto detection I'd be willing to consider choosing based on
char set, but I don't much like huristics beyond that.  Its
complicated, and will add a lot of complicated code.  Why don't we
just let the user tell us what language they'd like us to change to?
Humans are good at this sort of thing lets use that.  I have some
questions about the char set identification too, supose a user is
using an english language and we encounter chars that could be in
ukrainian or russian, witch do you switch to?


On Mon, Jul 19, 2010 at 09:23:42AM -0700, Bill Cox wrote:
On Mon, Jul 19, 2010 at 6:31 AM, Joanmarie Diggs
<joanmarie diggs gmail com> wrote:
Are you planing to implement the switching of languages in the synth
only or replacing orca's locale?

Can I have a third choice? <smile>

Based on my limited understanding of the problem, I suspect the best
solution would be to have "language profiles" in Orca, like described
in bug 419742, that users can switch between quickly.  In addition, it
would be great to have speech-dispatcher/opentts on the back end run a
simple heuristic to detect what language is actually being spoken, the
way voiceman does, and to temporarily switch when another language is
detected.  We could perhaps simply copy the heuristic.

If I understand correctly, there are many Russian web sites that have
a lot of English embedded, with no simple means of determining the
language, except to look at the characters used.  Since there is
little or no overlap between Russian and English letters, it's easy to
detect when to switch.  However, I've been told that it's harder to
detect the switch between Russian and Ukraine, and that users
currently manually switch in this case.

Perhaps we could do better language detection by loading the most
common 20,000 or so words per language profile, and matching each
phrase to be spoken to them.

orca-list mailing list
orca-list gnome org
Visit for more information on Orca.
The manual is at
The FAQ is at
Netiquette Guidelines are at
Log bugs and feature requests at
Find out how to help at

[Date Prev][Date Next]   [Thread Prev][Thread Next]   [Thread Index] [Date Index] [Author Index]