Databases Reference
In-Depth Information
as the so-called version log containing the different versions of ontology concepts
during their lifetime.
Changes are either basic (simple) or composite and defined declaratively using
the change definition language (CDL), which is based on RDF/OWL. Both kinds
of changes are determined by evaluating the old and new ontology versions w.r.t.
rule-based change definitions. For example, the change definition
8
p
2
P; A
2
C
W
addDomain
.p; A/
:
hasDomain
.p;A;v
i 1
/
^
hasDomain
.p;A;v
i
/
specifies that the basic change
addDomain
(
p, A
)toadd
A
as the domain of property
p
has occurred when this domain has not been in the old version
v
i 1
but in the
changed version
v
i
. Composite changes are more difficult to determine since they
involve several ontology elements that may be subject to changes themselves that
may have to be taken into account. The correct identification of such changes is
important to correctly adapt instances of the ontology. For instance, we may have
two basic changes to move property
p
from class
C1
to class
C2
followed by a
subclass addition between
C1
and
C2
. Treating these changes independently would
first delete all properties
p
in instances of
C1
. However, the following addition of a
subclass relationship between
C1
and
C2
would require the addition of property
p
to the
C1
instances. By finding out that the two basic changes realize the composite
change of moving up
p
in the class hierarchy, the unnecessary deletions of
p
values
can be avoided.
Detection of high-level changes in RDF/S ontologies
:
Papavassiliou et al.
(
2009
)
focuses on the detection of high-level changes (diff) between two RDF/S-based
ontology versions. Their framework uses a formal language to define changes and
distinguishes between basic, composite, and heuristic changes. Heuristic changes
refer to changes that are detected by matchers employing heuristic techniques to
determine that classes have been renamed, merged, or split. The proposed algorithm
focuses on the detection of basic and composite changes and utilizes the so-called
low-level delta containing the RDF triples that have been added and deleted between
two versions V1 and V2 of a RDF/S knowledge base. Changes are described by
triples consisting of (1) required added RDF triples, (2) required deleted RDF
triples, and (3) a set of conditions that need to be fulfilled. For instance, the change
Delete Superclass(x,y)
, which removes the is-a relationship between
x
and
y
, can
be described as follows: (1) no added triple exists, (2) the deletion of a triple (
x
, sub-
ClassOf,
y
) exists, and (3)
x
is a class in V1. The detection algorithm first uses the
low-level delta and the change descriptions to find potential changes between V1
and V2. The second step then iteratively selects changes that meet the conditions
and reduces the set of changes in the low-level delta. The algorithm first identifies
composite changes and then basic ones.
