Open dyamon opened 3 years ago
We decided to implement NI
s using the following rule
NI(X) <- X rsa:EquivTo Y, named(Y).
Note that removing NI
s entirely from Def. 4 and from the computation of the canonical model might have an impact also on the theoretical results.
We need to check that Theorem 3 remains true even when removing the NI
from Def. 4
This is just to keep track of the discussion about the instantiation of
NI
predicates in the RSA canonical model. In the paperNI
in mentioned multiple times.During the definition of canonical model for
O
:During the definition of filtering program:
Furthermore the predicate
named
is introduced in the filtering program seemingly with the same semantics.At the time of writing we agreed on the following:
named
is instantiated on all constants appearing in the original ontologyO
;NI
is defined as the set of constants inO
along with any fresh constant introduced during the canonical model computation that are equivalent to a constant inO
;NI
with the following rule:Despite this decision we are still not sure why the paper introduces
NI
during the definition of the logic program used to compute the canonical model, since its instantiation depend on the canonical model itself.