Dave Raggett | 30 Jul 14:29 2004
Picon

Re: [css3-speech] Comments: 'voice-volume'


-----BEGIN PGP SIGNED MESSAGE-----
Hash: SHA1

First, please accept my apologies for the delay my responding to
your comments.

In general, the CSS3 speech module is tied to the definition of the
W3C Speech Synthesis Markup Language (SSML), which is currently a
Proposed Recommendation, and this applies to the voice-volume
property amongst others. I have therefore added the www-voice list
to the thread, and maybe the creators of SSML can respond to the
detailed points you raise.

On Wed, 28 Jul 2004, csant wrote:

> * 'voice-volume'
> <numbers> are defined as a volume scale ranging from 0 to 100, and
> keywords are defined as a 6-point scale to cover the same range,
> but "the mapping of these to numerical volume levels is
> implementation dependent".
> 
> My take would rather be to match the 6-point scale to the numeric
> scale - mapping these values to the exact amplitude will be
> implementation dependent. In the most flexible case I might think
> of defining a keyword as covering a range of numeric values (say,
> e.g., 'medium' as suggested to be matched somewhere in the range
> between 50 and 70). I am sure there is a rationale behind defining
> two scales to cover the same range and leaving the mapping
> completely open.
(Continue reading)

Dave Raggett | 30 Jul 14:46 2004
Picon

Re: [css3-speech] Comments: 'pause-before', 'pause-after' and 'pause'


-----BEGIN PGP SIGNED MESSAGE-----
Hash: SHA1

On Wed, 28 Jul 2004, csant wrote:

> * 'pause-before', 'pause-after', and 'pause'
> The keyword values have now been added, indicating the prosodic
> strength of a pause.
> 
> I do not think the prosodic strength can be really measured in the
> pauses:  it is actually measured in the tone inflections of the
> voice before and after the actual pause. A long pause can be
> 'strong' or 'weak', and the real prosodic strength is given by how
> the last word before the pause is spoken.
> 
> I would see a neutral approach of 'x-short', 'short', 'medium',
> 'long' and 'x-long' more appropriate.

The pause properties are tied to the definition of the SSML break 
element, See:

  http://www.w3.org/TR/2004/PR-speech-synthesis-20040715/#S3.2.3

It notes: The break element is an empty element that controls the
pausing or other prosodic boundaries between words. I have added
www-voice to the thread, and perhaps the creators of SSML can
respond to your comments on prosodic boundaries.

> The problem of defining pause as being inserted "between the
(Continue reading)


Gmane