Re: non-entailment test for imports (was Re: Proposed response to Golbeck regarding imports issue)

On June 17, Jim Hendler writes:
> 
> At 2:29 PM +0100 6/17/03, Ian Horrocks wrote:
> >Jim,
> >
> >Quoting from Test:
> >
> >     An OWL Lite or OWL DL document D is consistent with respect to a
> >     datatype theory T if and only if there is some abstract OWL
> >     interpretation I with respect to T such that I satisfies an
> >     abstract ontology O equivalent to D, in which O has a separated
> >     vocabulary; (see [OWL Semantics and Abstract Syntax]).
> >
> >     An OWL Full document D is consistent with respect to a datatype
> >     theory T, if and only if there is some OWL Full interpretation I
> >     with respect to T such that I satisfies all the RDF graphs in some
> >     imports closed collection containing an RDF graph equivalent to D.
> >
> >     ...
> >
> >     An OWL consistency checker MUST be sound: it MUST return
> >     Consistent only when the input document is consistent and
> >     Inconsistent only when the input document is not consistent, with
> >     respect to the datatype theory of the checker.
> >
> 
> 
> I will answer this one in this WG (instead of moving to rdf-logic) 
> because I think it is VERY important to our design -- as I understand 
> our tests, the entailment tests MUST agree that something is 
> entailed.  However, it is my understanding that the non-entailment 
> tests do not require stating that a contradiction occurs, they may 
> return "Unknown"

Of course you can return "Unknown", but then you are incomplete.

> -- so if I want to pass the imports002 test but do 
> what Dan does, I could simply note that it includes a pointer to a 
> non-imported namespace and that the entailment occurs there,and 
> therefore I can imagine a tool producing a proof that essentially 
> says:
> 
> This document alone may not entail XXX, however it points to another 
> document that, if imported, would entail XXX.

Sure - it is perfectly fine to include this extra information (which
may or may not be used by a client application), but this doesn't
alter the fact that the basic answer must be "not entailed" or
"unknown" (and if "unknown", that the reasoner is incomplete).

> This first statement above (may not entail XXX) would mean it passed our test.

"unknown" would mean that it passed as an incomplete reasoner.

> 
> This would be within the letter and the law of our design (and maybe 
> even the spirit of the design for some of us).

There is no problem with a reasoner returning extra information,
provided that the actual answer is in accordance with the
specification in Test. If the tool doesn't claim to be an OWL
reasoner, but an OWL "indicator of what would be entailed if one were
to import xxx", then that is another question.


At this point some may want to point out that Test doesn't say
anything about entailment; it doesn't have to, because entailment is
trivially reducible to consistency. I.e., Ont1 entails Ont2 iff every
statement Si in Ont2 is entailed by Ont1, and Ont1 entails Si iff adding
the negation of Ai to Ont1 results in an inconsistent document. 

In the case of the imports002 test, adding the assertion that Socrates
is an instance of the complement of Mortal does *not* make premises002
inconsistent, so we have a negative entailment. In contrast, adding such
an assertion to premises001 *would* make it inconsistent because of
the additional imported information about man being a subclass of
mortal.

Ian

> If I am wrong on this, then I am very worried about our design of the 
> imports002 test, and will consider that this may be new evidence that 
> would convince me to ask Guus to reopen the imports issue.  If I'm 
> right, then I don't care if you think the above is a bad way to lead 
> my life, I just want to make sure that we have not ruled out 
> something that the chair believes the WG agreed not to rule out.
>   -JH
> 
> 
> (again
> 
> 
> 
> 
> >
> >A document containing only the assertion that x is of type
> >http://...foo#bar is clearly consistent according to the above
> >specification, regardless of the contents of http://...foo. If you
> >have a consistency checker that imports the contents of http://...foo
> >and returns "Inconsistent" if this leads to an inconsistency, then
> >your consistency checker is unsound according to the spec.
> >
> >
> >If you would like to have a tool that does some automatic importing,
> >then the obvious solution is to have it automatically extend the
> >ontology with the required import statements. Using this kind of tool
> >along with the existing imports statement lets everyone do more or
> >less what they want to do - you and Dan can do different kinds of
> >automatic importing, and others can do only explicit importing.
> >
> >Ian
> >
> >
> >
> >
> >
> >
> >On June 17, Jim Hendler writes:
> >>  At 11:53 PM -0400 6/16/03, Peter F. Patel-Schneider wrote:
> >>
> >>  >If there is not an imports statement, then OWL does not license importing.
> >>  >Any software that does so is not fully OWL compliant.  The negative
> >>  >entailment test Imports-002 is a test of this situation.
> >>  >
> >>  >
> >>  >peter
> >>
> >>
> >>  Peter - negative entailment test002 reads:
> >>
> >>  If a premise document uses a namespace but does not import the
> >>  document corresponding to the namespace, then the premises do not
> >>  necessarily entail anything that is entailed by the conjunction of
> >>  the two documents.
> >>
> >>
> >>  "The premises do not necessarily entail" which is not the same as
> >>  saying the premises necessarily do not entail.
> >>
> >>  Further, the definition of a non-entailment test is:
> >>
> >>  3.3. Non-Entailment Tests These tests use two documents. One is named
> >>  premisesNNN.rdf, the other is named nonconclusionsNNN.rdf. The
> >>  nonconclusions are not entailed by the premises. Such entailment is
> >>  defined by the OWL semantics [OWL Semantics and Abstract Syntax],
> >>  (see also OWL Full entailment). Exceptionally, test imports-002
> >>  includes a third document.
> >>
> >>
> >>  i.e. not entailed by, which I believe is not the same as "the
> >>  negation is entailed by"
> >>
> >>
> >>  I read this one carefully before I was willing to approve it, and
> >>  this statement is, indeed, consistent with the decision taken by the
> >>  group.
> >>  Our normative test document is therefore consistent with what I
> >>  described yesterday and I believe this still stands.
> >>
> >>    -JH
> >>
> >>
> >>  --
> >>  Professor James Hendler				  hendler@cs.umd.edu
> >>  Director, Semantic Web and Agent Technologies	  301-405-2696
> >>  Maryland Information and Network Dynamics Lab.	  301-405-6707 (Fax)
> >>  Univ of Maryland, College Park, MD 20742	  *** 240-277-3388 (Cell)
> >>  http://www.cs.umd.edu/users/hendler      *** NOTE CHANGED CELL 
> >>NUMBER ***<!doctype html public "-//W3C//DTD W3 HTML//EN">
> >>  <html><head><style type="text/css"><!--
> >>  blockquote, dl, ul, ol, li { padding-top: 0 ; padding-bottom: 0 }
> >>   --></style><title>Re: Proposed response to Golbeck regarding imports
> >>  iss</title></head><body>
> >>  <div>At 11:53 PM -0400 6/16/03, Peter F. Patel-Schneider wrote:</div>
> >>  <div><br></div>
> >>  <div>&gt;If there is not an imports statement, then OWL does not
> >>  license importing.<br>
> >>  &gt;Any software that does so is not fully OWL compliant.&nbsp; The
> >>  negative</div>
> >>  <div>&gt;entailment test Imports-002 is a test of this situation.<br>
> >>  &gt;</div>
> >>  <div>&gt;</div>
> >>  <div>&gt;peter</div>
> >>  <div><br></div>
> >>  <div><br></div>
> >>  <div>Peter - negative entailment test002 reads:</div>
> >>  <div><br></div>
> >>  <div>If a premise document uses a namespace but does not import the
> >>  document corresponding to the namespace, then the premises do not
> >>  necessarily entail anything that is entailed by the conjunction of the
> >>  two documents.</div>
> >>  <div><br></div>
> >>  <div><br></div>
> >>  <div>&quot;The premises do not necessarily entail&quot; which is not
> >>  the same as saying the premises necessarily do not entail.</div>
> >>  <div><br></div>
> >>  <div>Further, the definition of a non-entailment test is:</div>
> >>  <div><br></div>
> >>  <div>3.3. Non-Entailment Tests These tests use two documents. One is
> >>  named premisesNNN.rdf, the other is named nonconclusionsNNN.rdf. The
> >>  nonconclusions are not entailed by the premises. Such entailment is
> >>  defined by the OWL semantics [OWL Semantics and Abstract Syntax], (see
> >>  also OWL Full entailment). Exceptionally, test imports-002 includes a
> >>  third document.</div>
> >>  <div><br></div>
> >>  <div><br></div>
> >>  <div>i.e. not entailed by, which I believe is not the same as
> >>  &quot;the negation is entailed by&quot;</div>
> >>  <div><br></div>
> >>  <div><br></div>
> >>  <div>I read this one carefully before I was willing to approve it, and
> >>  this statement is, indeed, consistent with the decision taken by the
> >>  group.</div>
> >>  <div>Our normative test document is therefore consistent with what I
> >>  described yesterday and I believe this still stands.</div>
> >>  <div>&nbsp;</div>
> >>  <div>&nbsp;-JH</div>
> >>  <div><br></div>
> >>  <div><br></div>
> >>  <x-sigsep><pre>--
> >>  </pre></x-sigsep>
> >>  <div>Professor James Hendler<x-tab>
> >>  </x-tab><x-tab>&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;
> >>  </x-tab><x-tab>&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;
> >>  </x-tab><x-tab>&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp; </x-tab>&nbsp;
> >>  hendler@cs.umd.edu<br>
> >>  Director, Semantic Web and Agent
> >>  Technologies<x-tab>&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp; </x-tab>&nbsp;
> >>  301-405-2696<br>
> >>  Maryland Information and Network Dynamics Lab.<x-tab>&nbsp;&nbsp;&nbsp;
> >>  </x-tab>&nbsp; 301-405-6707 (Fax)<br>
> >>  Univ of Maryland, College Park, MD 20742<x-tab>&nbsp;&nbsp;&nbsp;
> >>  </x-tab>&nbsp; *** 240-277-3388 (Cell)<br>
> >>  http://www.cs.umd.edu/users/hendler&nbsp;&nbsp;&nbsp;&nbsp;&nbsp; ***
> >>  NOTE CHANGED CELL NUMBER ***</div>
> >>  </body>
> >>  </html>
> 
> -- 
> Professor James Hendler				  hendler@cs.umd.edu
> Director, Semantic Web and Agent Technologies	  301-405-2696
> Maryland Information and Network Dynamics Lab.	  301-405-6707 (Fax)
> Univ of Maryland, College Park, MD 20742	  *** 240-277-3388 (Cell)
> http://www.cs.umd.edu/users/hendler      *** NOTE CHANGED CELL NUMBER ***

Received on Saturday, 21 June 2003 09:49:28 UTC