In OWL-DL whether a role name represented an object or a datatype property was
inferred
globally on the basis of the entire ontology. (If it was not possible to infer
which type a property
was, then the difference did not make any difference to interpretation of the
ontology.) It was
possible to use a property in (for example) cardinality restrictions without
making it explicit
which type of property was used.
The current syntax specification requires that every use of a property
explicitly encode whether
the property should be treated as a datatype or as an object property. The
extra verbosity
required might not be considered an issue for the (already cumbersome) abstract
syntax, but it
prevents any terse alternate syntax from being converted to a valid OWL 1.1
fragment.
Manchester OWL Syntax, for example, lets one to write expressions along the
lines of
hasAddress atleast 1
but there are two possible translations of this expression to OWL 1.1:
ObjectMinCardinality(1 hasAddress)
and
DataMinCardinality(1 hasAddress)
This is a significant problem for interfaces which allow users to write OWL
fragments in isolation.
This issue is related to that of object/datatype property punning. If such
punning were
disallowed then there would be no need for such explicit typing.
Original issue reported on code.google.com by jlc...@gmail.com on 11 Jun 2007 at 2:59
Original issue reported on code.google.com by
jlc...@gmail.com
on 11 Jun 2007 at 2:59