captsolo weblog

This is a hard problem, because RDF in general doesn't contain any layout information, but RDFa needs layout information because the RDF statements are embedded in HTML layout.

Rendering RDF without domain information usually doesn't produce very good results. Therefore, I wonder a bit how useful such a service would be.

Maybe you can explain why you want to do this in the first place.

I am talking about a situation when we already have HTML and RDF, and want to fuse them together by using RDFa.

An example situation is FOAF and SIOC data which many sites currently expose as external metadata. It might take a bit before all these sites convert to RDFa but it would be good if services which can consume RDFa (e.g., SearchMonkey) could make use of this data right away.

Can you not encode arbitrary RDF as HTML+RDFa? Assuming that humans are only concerned with what is visible and machines are only concerned with statements extracted from the document (visibility of which is not a requirement as such) is there any "stakeholder" who cares if RDF statements are associated with a particular DOM node?

P.S. I am not talking here about a UI for visualizing RDF which may not have a good generic solution. This post is just about using a web page as a carrier of RDF data for consumption by RDFa-aware applications.

I think that's very interesting if the XHTML is generated from some semantic source. A related case that makes me interested in this (and that made me retrieve this page via Google) is the processing of semantic markup. In our setting, we have documents in a semantic markup language and want to do several things with them:

1. presenting them to humans (i.e. as XHTML)
2. extracting RDF structures from them (for queries)

We have implementations for (1) and (2) but are now starting to combine them. For active documents, which offer interactive services to the user who is browsing them, we need semantics in the XHTML rendered by (1), and an obvious approach is to embed this semantics as RDFa (see the JOBAD project for details). As, in our case, both the XHTML and the RDF are generated from the same source, and our implementations of (1) and (2) are technically similar (both done in XSLT), it is easily possible to combine them.

Thats a very intersting topic and I also wondered how to automatically relate external RDF data with DOM nodes via RDFa.

From my point of view it makes sense to let the RDF and HTML representation of certain data seperated, because they must not necessarily describe the data with the same detail level, and relate them only via few RDFa statemants. These RDFa statemants simply relate roughly DOM nodes with external resource description. Maybe this mixed approach would allow combining the advantages of external RDF descriptions (e.g. easy to parse, no layout issues) with the advantages of embedded RDF (e.g. positional information).

I wrote a related posting: http://www.clauwa.info/2009/01/09/embedded-versus-external-semantic-metadata/