owl inconsistencies

Hallo,
When inferencing there are three moments at which
inconsistencies can be detected:
1) before inferencing
2) during inferencing
3) after inferencing.
Before inferencing we have a set of formulae.
Take the triples:
(1)
:x owl:oneOf (:a :b :c).
:d a :x.
This is a clear inconsistency and should be detected
of course. 
However:
(2)
:c a owl:restriction; owl:onProperty :p;
owl:minCardinality "3"^^xsd:integer.
However before inferencing no member of the class
exists. But members of the class might be discovered
during inferencing.
Consider the situation after inferencing:
there is a set of formulae, a set of substitutions to
the query and a proof of the result. Applying the
susbtitutions to the set of formulae together with the
proof gives a new set of formulae (the proof is a
formula too). The found solution is valid when this
new set of formulae is a consistent set; the solution
is not valid when this set is not consistent. 
Thus it follows that the detection of inconsistencies
needs only to be done after inferencing, thereby
eliminating invalid answers. 
Of course, inconsistencies like (1) could be detected
before or during inferencing (but need not to) in
order to enhance the efficiency of the inference
process. 
Conclusion: inconsistencies that are irreparable(i.e.
they cannot go away by the inferencing process) should
be detected as soon as possible; inconsistencies that
can change should only be detected after the
inferencing process: these include certainly
restrictions with: owl:someValuesFrom;
owl:cardinality; owl:minCardinality;
owl:maxCardinality. 
Checking consistencies at the end of inferencing
should even allow to introduce "assert" and "retract"
primitives i.e. adding or deleting triples from the
formulae during inferencing.In that case, however, all
inconsistency handling needs to be done after
inferencing.



__________________________________
Do you Yahoo!?
Yahoo! Finance Tax Center - File online. File on time.
http://taxes.yahoo.com/filing.html

Received on Sunday, 21 March 2004 06:59:59 UTC