RE: [regrep-semantic] draft explanation of Semantic enhanced registry queries

["Jeff Pollock" <Jeff.Pollock@networkinference.com>]

as an extension to this email, it may be useful for folks to review the attached mini-whitepaper from NI about the different kinds of reasoning.
 
not intended to be a marketing ploy, disregard if it's not useful!
 
-Jeff-

	-----Original Message----- 
	From: ewallace@cme.nist.gov [mailto:ewallace@cme.nist.gov] 
	Sent: Tue 4/20/2004 5:21 PM 
	To: regrep-semantic@lists.oasis-open.org 
	Cc: 
	Subject: [regrep-semantic] draft explanation of Semantic enhanced registry queries
	
	


	SCM SC:
	
	Here is some material to chew on regarding mixing DL reasoning into
	reg/rep queries.  It is rough, not complete, and not in the form of
	use cases.  None-the-less it would help me toward my action item if
	I could get feedback on this.
	
	
	====================================================================
	
	
	Reasoning in support of Reg/Rep queries:
	
	If we were to store Semantic Content in OWL DL form, we would have the
	power of Description Logic classification available for use in
	querying both the Semantic Content itself as well as other RegRep
	content linked to that Semantic Content.  This latter bit will require
	some careful thought, though.
	
	DL classification is where a reasoner determines the following:
	
	1) which classes in an ontology are subclasses of a class specified
	by some complete description, that is - a description containing
	necessary and sufficient conditions for membership. Some examples of
	this from [1] are "Bus Drivers are Drivers" and "Drivers are Grown Ups."
	
	2) which instances can be classified as members of the class specified
	by a complete description. Some examples of this from [1] are "Pete is
	a Person, Spike is an Animal."
	
	Reasoning supporting type 1 classification is called TBox (Terminology Box)
	reasoning. Reasoning supporting type 2 classification is called ABox
	(Assertion Box) reasoning.  Some DL reasoners such as FaCT support
	only TBox reasoning* while others such as Racer support both types.
	
	I have been using the term "restriction" within SCM SC discusssions to
	refer to a complete description in OWL.  My assertion has been that we
	could compose DL classification with an SQL-style query by putting a
	restriction in place of one or more of the query parameters.
	
	If a registry query is of the form:
	 SELECT <attribute*>
	        FROM <extent>
	        WHERE <condition>
	        ORDER BY <attribute>
	
	Then a restriction (via reference to a definition in an OWL ontology
	or by description in place) could be used to define the extent for
	this query in place of something like a class/table name.  This would
	allow a registry user to choose his level of abstraction for a query
	or to specify an extent for the query which was not defined in any
	schema or ontology in the repository. (It should also be apparent that
	a restriction could be used in place of a WHERE clause as well.)
	
	An example of this which references OWL classes defined by the
	ontology in [1] might read:
	 SELECT firstName, lastName
	   FROM OWLrdf/xml <owl:equivalentClass rdf:resource="http://cohse.semanticweb.org/ontologies/people#grownup"/>
	   WHERE home = "Rockville"
	   ORDERED BY lastName
	
	An example of a restriction which could be used to retrieve duck
	instances would be:
	   FROM
	    <owl:equivalentClass>
	      <owl:Class>
	        <owl:intersectionOf rdf:parseType="Collection">
	
	          <owl:Restriction>
	            <owl:onProperty>
	              <owl:ObjectProperty rdf:about="#hasPart"/>
	            </owl:onProperty>           
	            <owl:someValuesFrom>
	              <owl:Class rdf:ID="Bill"/>
	            </owl:someValuesFrom>
	          </owl:Restriction>
	
	          <owl:Restriction>
	            <owl:onProperty rdf:about="#hasPart"/>
	            <owl:someValuesFrom>
	              <owl:Class rdf:ID="Tail"/>
	            </owl:someValuesFrom>
	          </owl:Restriction>
	
	          <owl:Restriction>
	            <owl:onProperty>
	              <owl:ObjectProperty rdf:ID="hasWalk"/>
	            </owl:onProperty>
	            <owl:hasValue>
	              <Walk rdf:ID="Waddle"/>
	            </owl:hasValue>
	          </owl:Restriction>
	
	          <owl:Restriction>
	            <owl:onProperty>
	              <owl:ObjectProperty rdf:ID="hasSound"/>
	            </owl:onProperty>
	            <owl:hasValue>
	              <Walk rdf:ID="Quack"/>
	            </owl:hasValue>
	          </owl:Restriction>
	
	        </owl:intersectionOf>
	      </owl:Class>
	    </owl:equivalentClass>
	
	
	* FaCT++ provides a function to simulate ABox reasoning in certain
	cases by redefining the instances as classes.
	
	[1] OWL Reasoning Examples
	http://owl.man.ac.uk/2003/why/latest/
	
	
	====================================================================
	
	-Evan

Copy of Tech_Whitepaper_Basic_DL_Reasoning.pdf