Re: Defining a pragmatic semantic description of SVG content.

I think, you forgot the desc element - such elements
are a source of good information too, if available and
a pretty good chance for authors to describe the
meaning, structure or useful interpretation of the
whole document, if desc elements are a direct children
of the svg document or the same for the related
group or element, if they are a direct parent of a
group or an element.

And I think, similar as the pair h+section in XHTML2,
from grouping, title, desc (and text elements) a
structure can be derived automatically to generate
a cascade of headings and sections especially for
screen readers and for an alternative text presentation
in general to offer some helpful structure of
such a document - indeed, an SVG element can have
a lot of structure. It is not completely silent, if authors
want them to have stucture.
Even if I do not know, how screenreaders present
the meaning of a title and section cascade in 
(X)HTML, there should be a 'simple transformation'
from SVG to (X)HTML to generate a simple structured
(X)HTML alternative from text containing elements 
to generate a useful representation in screenreaders.

One good approach for authors could be to have
mainly one title and one desc element at the 
beginning of the document as the main alternative
textual presentation. This should be possible for
every author, even if it will get to hard to some
applications and some authors, to have offer a
more complex representations, using title and
desc in other elements of the document too, which
would be a more advanced approach for some 
applications too. 

Note, that text elements can contain title and desc
too, therefore there is a requirement to destinguish
between the text content and the content from
title and desc. This needs a slightly different representation
to indicate what is content and what description.
It belongs to the same section in the cascade, but needs
to be separated form each other.

The SVG specification notes too, that 
'Description and title elements can contain marked-up text 
from other namespaces.'
Therefore authors can for example use XHTML to get
an addtional structure in title and desc, helping user agents
in general to generate a useful alternative text representation
of the SVG document.

The problem, that typically text elements in SVG do not tell on
the meaning of the presented text could be covered in the
future with the already discussed role attribute.
Especially for graphic formats like SVG it would be useful
to have a defined values list for a role attribute aligned along
the typical text elements with a semantical meaning defined in
HTML4 and the XHTML2 draft, maybe HTML5 too.
This would be a simple solution for authors to be able to
express the meaning of text in a structure SVG element 
containing larger parts of text. 
I have already some concrete poetry with and without animation,
currently using the class attribute for the purpose, to express the
semantical meaning of the text elements used in the document. 
This could aready help with an aural style sheet as an alternative
presentation. But with predefined values for a role attribute
of course it would be much more useful, even without stylesheets
and even if a user-agent does not care about aural stylesheets.

Received on Monday, 28 January 2008 13:57:40 UTC