W3C home > Mailing lists > Public > www-style@w3.org > January 2011

Re: [css3-speech] Phonemes: request for feedback

From: Daniel Weck <daniel.weck@gmail.com>
Date: Thu, 13 Jan 2011 14:26:08 +0000
Message-Id: <6F4DEADF-FB49-4885-A0C5-959B3CD04B4C@gmail.com>
To: www-style@w3.org
Given that HTML lacks the ability to let authors include pronunciation  
lexicons in their documents (for example, by referencing PLS files via  
some recognized link/rel/href scheme), document-wide pronunciation  
instructions would have to be provided by the user-agent, or more  
exactly by the web-browser's speech engine.
In practice, the scope of CSS3-Speech phonemes would be limited to a  
relatively small number of words in the document, such as for special  
cases that are known not to be handled correctly by TTS renderers.
It seems to me that the lack of support for document-wide  
pronunciation lexicons is a bigger problem (as far as authors are  
concerned) than the lack of pronunciation instructions based on CSS  
I could be wrong though, so thank you all for your feedback !
Regards, Daniel

On 13 Jan 2011, at 09:38, Andrew Cunningham wrote:

> Would having phoneme in CSS3 Speech be practical. It may work well
> with a small phrase or a few words, but when you need phonetic
> representations of multiple paragraphs of text would it be workable?
> Andrew
> On 13 January 2011 14:05, jtchen0901 <jtchen0901@itri.org.tw> wrote:
>> I kind of agree with the concerns of Charles.
>> The capability of inline usage of phonemes are desired,
>> Since the word 'speech' is in the title of 'CSS3 Speech', the  
>> phonemes
>> property sounds reasonable to be proposed in it.
>> I understand the suggestion to remove phoneme,
>> but, the existence of phonemes in CSS3 Speech seems not redundant.
>> Jiang-Chun Chen
>> --------------------------------------------------
>> 寄件者: "Belov, Charles" <Charles.Belov@sfmta.com>
>> 日期: 2010年12月14日 03:07
>> 收件者: "www-style list" <www-style@w3.org>
>> 副本: "Daniel Weck" <daniel.weck@gmail.com>
>> 主旨: RE: [css3-speech] Phonemes: request for feedback
>>> Daniel Weck wrote on Friday, December 10, 2010 8:22 PM
>>>> The "phonemes" property [1] described in the current Working
>>>> Draft of the CSS3 Speech Module covers functionality that is
>>>> arguably not related to styling. The concept of "phonemes" is
>>>> inherited from SSML [2], and indeed represents an important
>>>> aspect of speech synthesis.
>>>> However, the relevance of phonetic instructions in the
>>>> presentation layer needs to be discussed.
>>>> To bring concrete arguments into this discussion, I would
>>>> like to cite the ongoing revision of the DAISY open-standard
>>>> (format for Digital Talking Books and synchronized text/audio
>>>> publications). ANSI/NISO Z39.86-AI defines a markup grammar
>>>> [3] that 'adopts' SSML "phonemes"
>>>> to enable the provision of pronunciation instructions on XML
>>>> elements that are not within the SSML namespace. In other
>>>> words, this specification considers that phonetic
>>>> instructions are part of the data layer, not part of
>>>> styling/presentation. In this case, pronunciation information
>>>> is inlined within the content itself, but it could
>>>> alternatively be provided via some non-CSS out-of-band mechanism.
>>>> Additionally, the working group responsible for the ongoing
>>>> revision of the EPUB open-standard (the IDPF industry format
>>>> for electronic
>>>> publications) is also in the process of drafting the adoption
>>>> of SSML elements for inline use within XHTML markup [4].
>>>> Due to the contentious nature of the "phonemes" property in
>>>> the current CSS3-Speech Working Draft, and because of the
>>>> planned release schedule of both aforementioned specification
>>>> works, I would like to strongly urge interested parties to
>>>> provide feedback as soon as possible.
>>> I won't repeat my arguments from
>>> http://lists.w3.org/Archives/Public/www-style/2010Oct/0466.html as
>>> you've seen them.
>>> Nevertheless, if phonetics were to be eliminated from CSS, they
>>> would need to be added to HTML to permit inlined phonetics.
>>> Overloading the title tag with this feature would lead to
>>> potentially inappropriate title tags for sighted website
>>> visitors.  I say potentially, since phonetics in a title
>>> attribute may well be appropriate for some websites.
>>> I am also not clear upon reviewing the SSML spec how an HTML Web
>>> page would be able link to an SSML file the way HTML links to
>>> CSS.   From http://www.w3.org/TR/speech-synthesis/#S1:  "The
>>> markup may be produced either automatically, for instance via
>>> XSLT or CSS3 from an XHTML document, or by human authoring.
>>> Markup may be present within a complete SSML document (see
>>> Section 2.2.2) or as part of a fragment (see Section 2.2.1)
>>> embedded in another language, <em>although no interactions
>>> with other languages are specified as part of SSML
>>> itself.</em>" (Emphasis added)
>>> But this seems to imply, if I am reading this correctly, that
>>> SSML could be embedded in an HTML or CSS document.  I would
>>> hope I don't have to repeat the SSML code itself in each HTML
>>> document.
>>> I realize it is not CSS's concern whether phonemes are inlined
>>> in HTML or coded as SSML within or linked to HTML, but unless
>>> such capabilities exist or are planned, then my desire for
>>> phonetics within CSS have not been relieved.
>>> Hope this helps,
>>> Charles Belov
>>> SFMTA Webmaster
Received on Thursday, 13 January 2011 14:26:44 UTC

This archive was generated by hypermail 2.4.0 : Friday, 25 March 2022 10:07:54 UTC