XPath, binary content and RDF

All,

After surveying many of the folks here at W3C on how to best represent 
binary content and XPath-style querying in RDF...

So to the first question re: pointing to side chains using Xpath and 
doing similar work in RDF...if the CML spec has semantics (either 
explicitly described, or described by you) about which elements are side 
chains, then you should be able to take an xpath approach to those 
documents as you described.  However, as XPath does its business by 
connecting to the *structure* of the XML parse tree, and RDF is 
fundamentally structured differently, it would be interesting to port 
the xquery language into RDF.  The RDF Query group has asked for input 
into the utility of trying to do this, and I encourage all of you to 
think about such utility on this list.  One point:  it's very helpful  
that the semantics be explicit - the object to which the xpath refers is 
a side chain, not a piece of an xml parse tree.  I mention a tool below 
called GRDDL that helps extract some semantics from XML.

Now, how would we actually go about this in RDF? I would point everyone 
to Eric Prud'hommeaux's work on SQL-RDF interaction and representation, 
especially his paper at http://www.w3.org/2004/04/30-RDF-RDB-access/ 
which covers optimzed RDF access to relational databases, as well as a 
tool called Gleaning Resource Descriptions from Dialects of Languages or 
"GRDDL" (http://www.w3.org/2004/01/rdxh/spec) that takes "colloquial" 
XML (community-generated markup lang), runs it through XSLT and spits 
out RDF.  If you take the paper's approach and the GRDDL tools you can 
do some quick RDF/OWL representation of relational data and domain 
semantics without a lot of overhead.  I've got some example code that 
Eric has ginned up here (please be forgiving on details, we're not 
bioinformaticists quite yet!) that I'll post in a separate mail.

jtw

-- 
John Wilbanks
W3C Fellow
Semantic Web - Life Sciences
http://www.w3.org/People/all#wilbanks
wilbanks@w3.org
617-253-5845 (direct)
617-838-6333 (mobile; best voicemail #)
--

Received on Monday, 16 August 2004 19:41:07 UTC