- From: Harold Boley <boley@informatik.uni-kl.de>
- Date: Mon, 18 Sep 2000 19:53:52 +0200
- To: jos.deroo.jd@belgium.agfa.com
- CC: " - *www-rdf-logic@w3.org" <www-rdf-logic@w3.org>
Hi Jos,
> For the between example, would [usa between [canada pair mexico]]
> be a possible model?
Yes, this indeed nicely captures the USA focus in the "%"-comment of
between(mexico, usa, canada). % The USA is between Mexico and Canada.
and can also be extended for the symmetry of `between' in the first and
third arguments by just using a symmetric pair relation.
But someone could exploit the between(Lower, Middle, Upper) signature in
the south-to-north sense, grouping these countries differently such as
[[mexico south usa] and [usa south canada]], or even
[mexico south [usa south canada]],
or use some other permutation of nested triples. Also, `pair', `south',
`and', `argI' (from my previous email), and other new binary relations
can be avoided by just employing the above ternary `between' relation.
Moreover, for certain ternary relations there seems hardly any intuitive
binary reduction. Consider a purchase(Buyer, Object, Seller) relation.
For Horn relationships like
purchase(usa, alaska, russia). % The USA purchased Alaska from Russia.
there seems no better binary reduction than (in nested triple notation)
[purchase relship [[purchase1 arg1 usa] and
[[purchase1 arg2 alaska] and [purchase1 arg3 russia]]]]
To visualize this you can use a graph:
----------
| |
| |
| purchase |
| |
| |
----------
|
relship |
V
----------
| |
| |
------------| purchase1|------------
| | | |
| arg1 | | | arg3
| ---------- |
| | arg2 |
V V V
-------- -------- --------
| | | | | |
| | | | | |
| usa | | alaska | | russia |
| | | | | |
| | | | | |
-------- -------- --------
Or, for the original ternary relationship, again use a hypergraph:
-------- -------- --------
| | | | | |
| | | | | |
| usa | purchase | alaska | | russia |
| |--------------------------->| |
| | | | | |
-------- -------- --------
So, what would you think about an "N-ary RDF" where the triples in
"5. Formal Model for RDF" of http://www.w3.org/TR/REC-rdf-syntax/:
{pred, subj, obj} or pred(subj, obj)
are generalized to (N+1)-tuples:
{pred, arg1, ..., argN} or pred(arg1, ..., argN) ?
Best,
Harold
Received on Monday, 18 September 2000 13:54:25 UTC