W3C home > Mailing lists > Public > www-rdf-comments@w3.org > April to June 2003

Re: RDF Semantics: RDFS entailment lemma

From: pat hayes <phayes@ai.uwf.edu>
Date: Mon, 14 Apr 2003 14:27:13 -0500
Message-Id: <p05111b01bac0b5fed59f@[]>
To: herman.ter.horst@philips.com
Cc: www-rdf-comments@w3.org

>I agree that my two examples follow from the new rule rdfs12, and that
>this rule is valid.
>And I agree with Graham Klyne that the new rule rdfs7a follows
>from the old rule rdfs7a, so can be replaced by it.

Right, Graham was correct about that.

>  >But your example has some
>>other consequences: in fact, it entails that Resource is a subClass
>>of Class, ie that everything is a class.
>How do you obtain this?

Hmm, I thought I knew last night, but I cannot now reproduce that 
result. Never mind, pretend I never said it.

>  >
>>In order to prove the closure lemma, I need to somehow show that this
>>is the *only* way that the above entailment rule could possibly be

Because the entailment rule I mentioned

(x type A |- x type B) |- A subClass B

is in fact valid, and follows from the 'iff' in the semantic 
condition on subClassOf. So in order to show that this rule can be 
omitted from a complete rule base I have to show that any possible 
application of it is already covered by some other inference path. 
There is in fact a valid such rule for subPropertyOf as well, but no 
way it could possibly be invoked by an RDFS subproof, so that one can 
be safely omitted.

>>The only way I can see how to do this at present is by an
>>exhaustive analysis of the rule base, but I bet there is some elegant
>>way to do it which I don't have time to think of.
>>The general pragmatic conclusion seems to be that it is definitely
>>not a good idea to try to say  things about superproperties of
>>rdf:type, for sure :-) I propose to add the following paragraph as a
>>'warning' and also a brief commentary on this new rule:
>>The rule rdfs11

that should be rdfs12

>is a technicality, required in order to ensure the
>>truth of the following lemma. It is unlikely to be used in practice,
>>and will normally only produce redundant inference paths for some
>>items in the closure. In general, the property rdf:type is best
>>considered to be part of the logical machinery; as this rule
>>illustrates, imposing gratuitous conditions on rdf:type can produce
>>unexpected entailments. Similar strange conclusions can arise from
>>asserting that rdfs:Resource is a subclass of another class, for
>>example, or asserting unintuitive properties of rdfs:Class.
>I'm not sure whether these last two paragraphs are justified.
>Couldn't you also say that the new rule rdfs12 shows that the
>rdfs does not enable one to make the domain of rdf:type
>or any of its superproperties any smaller than it is
>(i.e., rdfs:Resource) by adding other domain statements?

That might indeed be a better way of putting it, thanks for the 
suggestion. I will re-draft this paragraph.

Thanks again for your help.

IHMC					(850)434 8903 or (650)494 3973   home
40 South Alcaniz St.			(850)202 4416   office
Pensacola              			(850)202 4440   fax
FL 32501           				(850)291 0667    cell
phayes@ai.uwf.edu	          http://www.coginst.uwf.edu/~phayes
s.pam@ai.uwf.edu   for spam
Received on Monday, 14 April 2003 15:27:16 UTC

This archive was generated by hypermail 2.4.0 : Friday, 17 January 2020 22:44:02 UTC