xml-dev - RDF/XML, Node vs. Edge labeled graphs was: Re: Services-basedautomation

[Jonathan Borden <jborden@mediaone.net>]

Thomas B. Passin wrote:
> Jonathan Borden wrote -
>
> >
> > The XML model defines a node labelled directed graph. In
> this model, arcs or
> > edges have the type "element" "attribute" "CDATA section"
> "comment" etc.
> >
> > The RDF model defines an edge labelled directed graph, for
> example arcs may
> > be labelled "color", "type.of.cheese" "type.of.sauce"
> "topping".

>
> Do the two models really represent anything different? Not
> really.  But if connections - a form of structure - are the
> most important thing to you, edge-labeled models might seem
> more attractive or "natural".

Right, so a starter for RDF is that the edge labelled model (i.e. triples),
layered on top of XMLs node labelled model (e.g. DOM,SAX,Infoset) , is a
more 'natural' way to represent semantic information.

But we can extract RDF from colloquial XML and serialize RDF as XML by
defining a mapping between the two representations.

I've written an XSLT RDF extractor for arbitrary XML
http://www.openhealth.org/RDF/rdfExtractify.xsl which results in a series of
rdf:Statement's i.e. a triple (predicate,subject,object). A predicate is a
URI which names the edge (i.e. an xlink:arcrole), a subject is a URI which
names the "from" node and the object is either a URI which names the "to"
node or a string which is a value. This extractor implements the "Harvesting
RDF Statements from XLink" note http://www.w3.org/TR/xlink2rdf/ .

One of the properties of 'flattening' or reifying a piece of XML into
triples is that a set of triples neatly fits into a relational table. R.V.
Guha has written an RDF database which operates on such triples
(http://web.guha.com/rdfdb/) and includes a SQL like query language.

Jonathan Borden
The Open Healthcare Group
http://www.openhealth.org