In “What I am Working On“, Stephen Downes beautifully explains why resource descriptions are not the adequate approach for supporting human learning and knowledge.
“A piece of knowledge isn’t a description of something, it is a way of relating to something.” … “this is why the semantic web is on the verge of becoming a very expensive failure”
Contrasting descriptions and relations, is very plausible. RDF’s triple descriptions of a Subject by a Predicate and an Object, are
- not designed for “statements of the type, ‘people like P thought that resources like R were rated Q’.”,
- and not for vague, informal text snippets, or utterances like “R1 is somehow similar to R2”,
- or even generic associations.
What RDF ontologies can handle well, is the low hanging fruit type of stale, encyclopedic “knowledge” such as “The Elephant is gray and :lives-in Africa” that is indeed a description, where the only “rating” person P is an abstract, omniscient authority.
But there are also people working on more humanized RDF approaches. The Nepomuk group’s CDS (conceptual data structures) allows for very vague and informal knowledge articulation. For instance, text snippets in a semantic wiki that are bullet points of the same unnumbered list, have only the vague relationship “before/ after”, but still, they have something in common.
Now even such informal wiki snippets may be incorporated into RDF, following this model (which originates from the task of a wiki interchange format, where each text snippet has to be addressable, anyway). See slide 10 in the rather dense presentation, where it is simplified using the elephant example.
Basic CDS relationships such as lists can now be expressed as “hasBefore” and “hasAfter”, and can model things like the above “resources that P rated Q”. And even the arbitrary associations can be expressed: using “hasLinkSource” and “hasLinkTarget”.
I think this is a very promising approach to make RDF usage more human.