Databases Reference
In-Depth Information
changes and the addition/deletion of definitions. More complex changes such
as merges or splits of concepts are not supported.
The detection algorithm is inspired by the UNIX diff operation but uses the ontol-
ogy graph structure and RDF triples
<
subject, predicate, object
>
as the basis for the
version comparison. Change detection between two graphs is based on IF-THEN
rules that specify conditions on triples in the old/new ontology and produce resulting
changes if the conditions are fulfilled. The authors argue that they can specify and
detect almost every change type using this mechanism except identifier changes.
O
ntology
e
volution e
x
plorer (OnEX) is a web-based system for exploring chan-
ges in numerous life science ontologies (
Hartung et al. 2009
). It uses existing
ontology versions and identifies the differences between succeeding versions of an
ontology. The differences are represented by evolution mappings consisting of sim-
ple changes (adds, deletes, updates of concepts/relationships, and attributes) that are
identified by comparing the unambiguous accession numbers of elements available
in life science ontologies (
Hartung et al. 2008
). OnEX can be used to determine
the stability and specific change history of ontologies and selected concepts of
interest. Furthermore, one can determine whether given annotations referring to an
ontology version have been invalidated, e.g., due to deletes. Such annotations can
then be semi-automatically migrated to be consistent with the newest version of the
respective ontology.
OnEX uses a tailored storage model to efficiently store all ontology versions in
its repository by utilizing that succeeding ontology version differ only to a small
degree (
Kirsten et al. 2009
). Currently, OnEX provides access to about 700 versions
of 16 life science ontologies.
The ontology diff algorithm proposed in
Hartung et al.
(
2010
) determines an
evolution mapping between two ontology versions. The evolution mapping consists
of a set of simple as well as complex ontology changes (e.g., merging or splitting of
concepts). The approach is based on an initial matching of the ontology version and
applies so-called Change Operation Generating Rules (COG rules) for deriving the
change operations of the evolution mapping. For instance, the rule for determining
a merge of multiple concepts looks as follows:
9
mapC
.a; c/
^9
mapC
.b; c/
^:9
mapC
.a; d /
^:9
mapC
.b; e/
^
a
¤
b
^
c
¤
d
^
c
¤
e
!
create
Œ
merge
.
f
a
g
;c/;
create
Œ
merge
.
f
b
g
;c/
The rule derives that concepts a and b are merged into concept c if there are two
match correspondences
mapC
(
a,c
)and
mapC
(
b,c
)andif
a
and
b
are not con-
nected to any other concept. The approach could be validated for different kinds
of ontologies.
Change detection using a version log
: Plessers and
De Troyer
(
2005
) builds
upon the KAON ontology evolution process (
Stojanovic et al. 2002
). The pro-
posed evolution process consists of five phases: (1) Change Request, (2) Change
Implementation, (3) Change Detection, (4) Change Recovery, and (5) Change Prop-
agation. The main difference is in the Change Detection phase where additional
implicit changes are detected based on the history (log) of previous changes as well
