RE: RDF [was: Proposed changes]

> > A potentially cool bit is that I think it should be possible to
> do this on
> > the fly : an RDF query could ask for all links in the content
> of that entry,
> > and the RDF store could get as far as <content> but then an XPath query
> > could be carried out on that, the results expressed in the RDF model.
>
> How would that work?  It seems to me that if you solve that, you have
> essentially eliminated the need for a special RDF/XML syntax - i.e., any
> XML could be directly consumed by an RDF application.

This is essentially along the lines of Uche's statement [1] "There is no
syntax" -
[[
The trick is to extract key metadata from XML documents -- or even
non-markup formats such as RDBMS -- and synchronize it into RDF models. You
can then treat these as a localized semantic Web.
]]

This only works when you have prior knowledge of the XML format. If you have
have that, you can usually make a direct mapping between that XML and the
RDF model. The weak link is when you start mixing stuff up from different
namespaces - the mapping might no longer hold, and you'd have to define it
not only for every language, but for every combination of languages. Using
RDF/XML ensures there's always a mapping to the RDF model.

> > Why not do it all as XML/XPath? Main reason being that the XML tree
> > structure doesn't match the web's general directed graph
> structure, probably
> > a more immediate practical problem being that the subtrees aren't very
> > portable across systems (merging isn't straightforward).
>
> It seems to me that the web's general structure is one where links tend
> to be embedded in mixed content.

Looking up from the source, yes, but from a few 1,000 feet it looks like
objects (documents) that are linked together.

> > some related notes : http://dannyayers.com/archives/001981.html
>
> Case in point: the above is an XML document with a number of links to
> your own site, and one to Simon Willison's site.

Right, and there is a disconnect between the semantics within the content
and any explicit semantics expressed outside. For the (RDF) Semantic Web the
links are latent semantics. As well as the need to get the 'blatant'
metadata systems sorted out (RDF storage, RDF/OWL inference etc), I
personally believe it's necessary to bridge to the latent semantics -
metadata hidden in XML and things like text search (Bayesian categorizing
etc, to infinity and beyond...). The blatant metadata systems are in the
(early) implementation phase, but the bridging to latent metadata still
hasn't been covered systematically. Fortunately I reckon most of the
groundwork has already been done (e.g. XPath), so it's only matter of gluing
the pieces together.

Cheers,
Danny.

[1] http://www-106.ibm.com/developerworks/xml/library/x-think12.html

Received on Sunday, 26 October 2003 05:31:29 UTC