Re: RDF graph representation

Dimitri Glazkov wrote:

> This may be a bit of sophistry, but since predicate may be used as
> subject (and object, I assume) in RDF, doesn't this make a
> node-arc-node representation of a triple invalid?
> 

To play devil's advocate 'no'

> For instance, how would you represent an OWL property restriction in
> such a graph mode? Wouldn't this mean that you'd have to point an arc
> into the middle of another arc?
> 
> Or am I missing something here? (which is probably the case) 
> 
> Illuminate me please..
> 

Consider a graph

eg:a eg:b eg:c .
eg:b eg:a eg:c .

OK so far?

So, the above representation *is* adequate since you are managing to 
consider it.

We can draw the traditional diagram of this graph.
It conveys the same information, and we can reconstitute the triples 
from the picture. Thus, the representation is adequate.

Obviously, when trying to understand what it might mean, you have to 
notice that the eg:b labelling one of the edges and the eg:b labelling a 
node is the same. But that goes with many visual representations. e.g. 
if you have an atlas, and the map of Eurasia is split over a number of 
pages, with Berlin in two of them, you have to notice that this is the 
same place because the two visual representations are labelled the same, 
despite one being shown along with Paris, and the other being shown 
along with Warsaw.

Visual representations *do* require the reader to do some work. The 
traditional visual representation of the RDF graph, specifically 
requires the user to notice when edge labels are the same (and when they 
are the same as a node label).

Traditionally, edge labels are referred to as colours in graph theory. 
The RDF graph is peculiar in having both coloured arcs and coloured 
nodes (partially). But (the non-colour blind) would not see it as 
strange being asked to notice that the blue edges were the same colour 
as a blue node.

Jeremy

Received on Friday, 23 July 2004 09:54:17 UTC