Databases Reference
In-Depth Information
the problem of taking two ontologies and combining them to form a third ontology.
So, for example, alignment of the ontologies in Fig.
8.2
might tell us that element
pairs with the same name are aligned with each other, that “Ugrad” matches with
“Undergrad,” and that “Office” matches with “Location.” Merging the two ontolo-
gies would result in a single ontology with the same structure as both ontologies
A
and
B
since the two are structurally identical, but there would be some resolution of
the naming differences (e.g., a decision would have to be made as to what to call the
node that describes undergraduates).
SMART is used for both merging and alignment. The algorithm is not completely
automatic for either operation; there are stages that must be performed by the user,
even after the initial setup has occurred.
SMART keeps track of its state with two data structures: the
Conflicts list
and
the
ToDo list
. The Conflicts list details the inconsistencies in the current state of the
process that must be resolved before the resulting ontology can be in a “logically
consistent state.” The ToDo list keeps track of operations which should be done but
are not required in order for the resulting ontology to be in a logically consistent state
(e.g., if an action results in two attributes in the same classes with a similar name,
SMART might add to the ToDo list a suggestion that one of them be removed). Since
determining the source of a problem may enable the user to optimize the problem's
resolution, each item in the ToDo and Conflicts list contains a reference back to the
operation that triggered it. More details of SMART and Prompt, particularly on the
matching and alignment aspects, can be found in
Falconer and Noy
[
2011
].
An outline of the SMART algorithm for merging is shown below. Note that the
merging process requires also performing an alignment, so steps for both appear in
the algorithm:
1.
The user performs setup by loading the ontologies,
A
and
B
and specifying some
options such as specifying if there is a preferred ontology.The result, the ontology
C
, is initialized to be a new ontology with a new root and
A
and
B
as that root's
children.
2.
SMART generates an initial list of suggestions of what should be aligned/merged.
In this stage, SMART relies largely on content or syntactic information. The
names of the objects are examined, but structural information (i.e., the position
of the classes or their participation in specific relations) is not used.
For each pair of classes
a
B
with identical names SMART either
merges the
a
and
b
in
C
or removes either
a
or
b
from
C
.
2
A
and
b
2
For each pair of classes
a
B
with linguistically similar names a
link is created between them in
C
. This means that both
a
and
b
are still in
C
,
but SMART suggests that they may need to be merged by adding them to the
ToDo list.
2
A
and
b
2
3.
The user selects and performs an operation such as merging a class or resolving
an item on the ToDo or Conflict lists.
4.
SMART performs any automatic updates that it can and create new suggestions.
It has the ability to:
