RE: LANG: need to CLOSE Issue 5.6 Imports as magic syntax

I am strongly in agreement with Jeff regarding the importance of
imports.

> The problem with this operational definition is that it could be
> interpreted in two ways: 

> 1) It specifies a procedure that all OWL systems must follow.

I don't think it does.  When I said

> > The containing ontology is extended by the content
> > of the imported ontologies.  

I explicitly intended to use the weasle words 'extended by'.  Which
could mean a lot of things. More to the point there is no requirement
that an implementation do anything.  Failure is acceptable and its
effect is implementation defined.

And the Texas example is a red herring, at least as regards to how I
intended it.  The point of my Texas example was that it referred to
locations in other states WITHOUT explicitly importing the ontologies
they were contained in.  What should be done with such references is
up to the various tools.  This is where I think we should provide the
latitude you are looking for, in the absence of explicit imports.

When an explicit imports is present, I expect the indicated ontology
to be imported, if possible.  But I don't know how we can give a
semantics to a relation that may easily be meaningless (the document
name is misspelled).  

I think of your statement

> A imports B means
> if B entails P then A entails P

as being more precisely stated as

  An occurence of 'A imports B' in (ontology A) means
  if (ontology B) entails P then (ontology A) entails P

But the only way to determine the value of (ontology B) is
operationally.  You have to dereference the URL.  That operation may
fail.  In which case (ontology B) is the empty set and the statment
above holds.

- Mike

-----Original Message-----
From: Jeff Heflin [mailto:heflin@cse.lehigh.edu]
Sent: Wednesday, October 30, 2002 8:53 AM
To: Smith, Michael K
Cc: webont
Subject: Re: LANG: need to CLOSE Issue 5.6 Imports as magic syntax

"Smith, Michael K" wrote:
> 
> 2 more cents.
> 
> > 3) The imports triples are considered extra-logical, and any statements
> > that contain owl:imports as a subject or object are undefined.
> > Furthermore, any imports statements that have a resource other than the
> > containing document as a subject are undefined.
> 
> OK.
> 
> > 2) The semantics essentially be "A imports B means if B entails P then A
> > entails P." ...
> 
> Imports is extra-logical by 3.  It has no formal semantics.

Yes, I shouldn't have used the term extra-logical there. Instead, I
should have said: "Certain triples that have imports as a subject or
object, or that have a resource other than the containing document as
subject do not have semantics.

> I know what you are trying to get at, I just don't think we have
> figured out how to say it.  And I don't know how to say it except
> operationally.  We all agree (?) that if importing B works (because
> all of my internet connections are working, the document B exists,
> the server it resides on is up, my firewall and anti-virus
> software are happy, ..., and A gets augmented with B) then the
> resulting KB will entail P.
> 
> I would be inclined to a statement like:
> 
>  The owl:imports statement is extra-logical.  It has an operational
>  interpretation.  The containing ontology is extended by the content
>  of the imported ontologies.  This extension is performed recursively
>  until there are no more imports to apply.  This attempt to extend
>  the ontology may not always be completely successful.  The effect of
>  such failure is implementation defined.
> 
>  If the interpretation of 'A owl:import B' is successful, then the
>  entailments derivable from the resulting KB are those derivable from
>  the union of the contents of A and B.
> 
> - Mike

The problem with this operational definition is that it could be
interpreted in two ways: 

1) It specifies a procedure that all OWL systems must follow. However,
this means that in the Texas ontology example which imports bordering
states, every application must always import every contiguous state,
even if it is only answering queries that concern Texas. I don't think I
could live with this interpretation.

2) The operation only specifies semantics, and any implementation that
comes up with the same answers as would be derived by this procedure is
correct. This is essentially equivalent to my entailment-based
definition, but much less formal. It does have the advantage that it is
clear that you cannot infer imports triples. I could potentially live
with this with some wordsmithing, but I think it makes our semantics
much more complicated because we now have a strange hybrid of
operational and model theoretic semantics.

Jeff

-----Original Message-----
From: Jeff Heflin [mailto:heflin@cse.lehigh.edu]
Sent: Wednesday, October 30, 2002 8:53 AM
To: Smith, Michael K
Cc: webont
Subject: Re: LANG: need to CLOSE Issue 5.6 Imports as magic syntax



"Smith, Michael K" wrote:
> 
> 2 more cents.
> 
> > 3) The imports triples are considered extra-logical, and any statements
> > that contain owl:imports as a subject or object are undefined.
> > Furthermore, any imports statements that have a resource other than the
> > containing document as a subject are undefined.
> 
> OK.
> 
> > 2) The semantics essentially be "A imports B means if B entails P then A
> > entails P." ...
> 
> Imports is extra-logical by 3.  It has no formal semantics.

Yes, I shouldn't have used the term extra-logical there. Instead, I
should have said: "Certain triples that have imports as a subject or
object, or that have a resource other than the containing document as
subject do not have semantics.

> I know what you are trying to get at, I just don't think we have
> figured out how to say it.  And I don't know how to say it except
> operationally.  We all agree (?) that if importing B works (because
> all of my internet connections are working, the document B exists,
> the server it resides on is up, my firewall and anti-virus
> software are happy, ..., and A gets augmented with B) then the
> resulting KB will entail P.
> 
> I would be inclined to a statement like:
> 
>  The owl:imports statement is extra-logical.  It has an operational
>  interpretation.  The containing ontology is extended by the content
>  of the imported ontologies.  This extension is performed recursively
>  until there are no more imports to apply.  This attempt to extend
>  the ontology may not always be completely successful.  The effect of
>  such failure is implementation defined.
> 
>  If the interpretation of 'A owl:import B' is successful, then the
>  entailments derivable from the resulting KB are those derivable from
>  the union of the contents of A and B.
> 
> - Mike

The problem with this operational definition is that it could be
interpreted in two ways: 

1) It specifies a procedure that all OWL systems must follow. However,
this means that in the Texas ontology example which imports bordering
states, every application must always import every contiguous state,
even if it is only answering queries that concern Texas. I don't think I
could live with this interpretation.

2) The operation only specifies semantics, and any implementation that
comes up with the same answers as would be derived by this procedure is
correct. This is essentially equivalent to my entailment-based
definition, but much less formal. It does have the advantage that it is
clear that you cannot infer imports triples. I could potentially live
with this with some wordsmithing, but I think it makes our semantics
much more complicated because we now have a strange hybrid of
operational and model theoretic semantics.

Jeff

Received on Wednesday, 30 October 2002 10:37:51 UTC