Re: A Simple Extension for RDFa Table Support

Ted, David,

I like this idea... just like I lot a number of other ideas that take
specific semantic knowledge of the HTML DOM and try to extract that
meaning into RDF. In my opinion, it's difficult to capture all of those
in a default RDFa specification.

I'd like RDFa to eventually support what I've called hGRDDL [1],
effectively in-place transformations of HTML and XHTML to take into
account dialect-specific semantics. One could specify these in-place
transformations in the <head>, almost the same way GRDDL transformations
are indicated.

Would you be interested in prototyping this as an hGRDDL transform? My
(well, it's also Jeni's and Nathan's now) RDFa parser is mostly set up
to do hGRDDL, if you hook into the right places [2].

-Ben

[1] http://portal.acm.org/citation.cfm?id=1346366.1346691
[2] http://www.w3.org/2006/07/SWD/RDFa/impl/js/rdfa.js

Edward Benson wrote:
> Hi everyone,
> 
> Our research group has been building several wiki- and blog-based tools for
> which RDFa support would be a great fit. While looking at how we might
> integrate it, we came upon a suggestion for how RDFa might achieve a cleaner
> representation when used with tables.
> 
> The fact that the <COL /> element of a table doesn't actually contain
> anything makes it impossible (under the current rules) to stash meaningful
> information there from the RDFa perspective. This means that we can put
> structured information across each row of a table, but not down each column,
> leaving us to repeat a lot of RDFa annotation information across each row.
> Here is a pseudocode example:
> 
> <TABLE>
>   <TR><TH>First</TH><TH>Last</TH></TR>
>   <TR subject="tj"><TD property="first">Thomas</TD><TD
> property="last">Jefferson</TD></TR>
>   <TR subject="ja"><TD property="first">John</TD><TD
> property="last">Adams</TD></TR>
> </TABLE>
> 
> What if instead, when parsing the DOM to extract structured information,
> whenever a <TD> element is encountered, the corresponding <COL> element is
> treated as if it were the DOM parent of the <TR> element for that cell. This
> would allow us to stash information both on the rows *and* the columns of
> the table, allowing for a much more compact representation:
> 
> <TABLE>
>   <COL property="first" /><COL property="last" />
>   <TR><TH>First</TH><TH>Last</TH></TR>
>   <TR subject="tj"><TD>Thomas</TD><TD>Jefferson</TD></TR>
>   <TR subject="ja"><TD>John</TD><TD>Adams</TD></TR>
> </TABLE>
> 
> Let us know what you think. It makes tables fit much more naturally, from
> our personal aesthetic point of view. As such, it may make it much easier
> for template writers (think blog themes, Wikipedia Info Boxes, etc) to
> incorporate into their templates and content.
> 
> Best,
> Ted Benson & David Karger
> 
> MIT CSAIL Haystack
> http://haystack.csail.mit.edu
>

Received on Wednesday, 10 June 2009 23:15:38 UTC