A proposal for entailment tests

Pat,

This is about
[[
  [14:48:35] scribe-Aaron
  -action Jos / write up something to describe these entailment
  tests (with help from Pat)
]] -- http://ilrt.org/discovery/chatlogs/rdfcore/2001-09-21.html#T14-48-35

I've tried to put something together and used already some
of your words...

--
Jos

A proposal for entailment tests.
-------------------------------

We have to make a distinction between entailment in
RDF [1] and entailment in RDFS [2].
In pure RDF, e.g. 'rdfs:range' and 'rdf:Property' are just
urirefs with no special meaning, but they do have a special
meaning in RDFS.  That is why we have to have extra rules,
or something, to capture all of RDFS.

We describe an RDF/RDFS entailment test in RDF (a matter
of eating your own dogfood) so that they have a precise
and machine understandable description
e.g.
  [ tc:graph g1, g2, g3 ] tc:entailrdf [ tc:graph g4 ].
describes the RDF entailment of the graph g4 given the
graphs g1, g2 and g3
and
  [ tc:graph g1, g2, g3 ] tc:rdfsentail [ tc:graph g4 ].
describes the RDFS entailment of the graph g4 given the
graphs g1, g2, g3 and the rules in [2].

<comment>
  tc: is a namespace prefix for a testcase schema
  gi is a uriref of a .rdf or .nt testcase graph
  we can write that straightforward in N-triples
</comment>

The syntax testcases can be described in the same
manifest file as
e.g.
  [ tc:graph g1 ] tc:entailrdf [ tc:graph g2 ].
  [ tc:graph g2 ] tc:entailrdf [ tc:graph g1 ].

[1] http://www.w3.org/TR/RDF-model-theory/#entailrdf
[2] http://www.w3.org/TR/RDF-model-theory/#rdfsentail

Received on Friday, 21 September 2001 19:14:25 UTC