Re: [orca-list] Capital, Capital, Capital

["Bohdan R. Rau" <ethanak polip com>]

Dnia 2008-03-18, wto o godzinie 12:25 +0100, Tomas Cerha pisze:
[...]
> Thanks everyone for the feedback.  I would like to ask those who hear
> the unwanted "capital" announcements to check whether they have the
> option DefaultCapLetRecognition in their Speech Dispatcher configuration
> set to "none" (should be the default).  If that's the case, I suspect
> the problem might be in the Speech Dispatcher's espeak module, since I
> believe this setting should be honoured and if it is not, it may be a
> bug.

Short test.

Gnome-terminal. While typing, Orca changes pitch for capital letters and
speaks character (without 'capital'). But when I try to insert character
with special key sequence (shift_right-alt_= a for 'Ä' on my keyboard)
Orca speaks nothing. While editing text in gedit, Orca says 'a', not
'Ä'. Accented letters coming from native key combination (right-alt_a
for 'Ä' on my keyboard) are spoken properly.

While moving with left/right, Orca also changes pitch for capitals, but
speaks one-character string, not character (it's very simple to check in
Polish, because string "w" and character 'w' are pronounced differently,
same for "z" and 'z'). Also for 'Å' I hear proper 'si', but for 'Ã' a
little strange 'o acute' instead of 'u kreskowane' (Polish special name
for 'Ã', which is known by eSpeak). Same for gedit.

In Firefox when moving by character, Orca sends (probably)
single-character string instead of character. So it's not possible to
distinguish between "Ã" and "u" (same pronunciation in Polish).


In flat review mode, Orca changes pitch for capitals, but also says
"capital a" for 'A' (in fact eSpeak gets:
<say-as interpret-as="tts:char">A</say-as>). Accented letters are spoken
properly, but "military spelling" is completely unusable.

For DefaultCapLetRecognition commented out and set to "none" results are
same.

So we can assume:

a) there is bug in speech-dispatcher espeak module
b) there is more than one bug in Orca causing problems with speaking
single characters.

But IMO if Orca (or other speaking applications) developers provides
it's own mechanism for announcing capital letters, they must not assume
the synthesizer has glass orb. If even this bug in speechd will be
corrected, and Orca will allow to choice capitals announcing mode -
there still will be problem, because user preferences will be stored in
two different places (Orca preferences and synthesizer configuration).
If there will be more than one user on single machine (each with it's
own preferences) - there will be very big problem.


> Yes, this is one of the major concerns which Speech Dispatcher tries to
> address and this is also one more reason for having a chance to bypass
> Orca's own verbalization.

Exactly. As for me, even "military spelling" should be moved to
synthesizer - may be something like <say-as interpret-as="tts:spell">?

Bohdan