Databases Reference
In-Depth Information
linguistic matching including one that uses a virtual document per element consist-
ing of the name, comments, and instance values of the element. Rimom is among
the best performing prototypes in the OAEI contests until 2009.
4.2.4
Asmov
Automated semantic matching of ontologies with verification (ASMOV) proto-
type (
Jean-Mary et al. 2009
) is among the best performing systems at the recent
OAEI match contests. Its most distinctive feature is an extensive postprocess-
ing of the combined matcher results to eliminate potential inconsistencies among
the set of candidate correspondences. Five different kinds of inconsistencies are
checked including the avoidance of the so-called crisscross correspondences, e.g.,
to prevent that for a correspondence between classes
c1
and
c1
0
, there is another
correspondence mapping - a child of
c1
to a parent of
c1
0
.
4.2.5
AgreementMaker
This ontology matching prototype is developed at the University of Illinois at
Chicago (
Cruz et al. 2009
). It provides a sophisticated GUI so that the user can
control the iterative execution of matchers. AgreementMaker was among the best
performing systems in the OAEI 2009 contest.
4.2.6
Harmony
Harmony is the match component within the Open Information Integration project on
developing a publicly available infrastructure for information integration (
Seligman
et al.
2010
). It provides many of the known features of previous match prototypes
as well as a GUI. Instance-based matching is not yet supported. The combination of
matcher results uses a nonlinear combination of similarity values to favor matchers
with higher similarity values (
Mork et al. 2008
) as briefly discussed in Sect.
3.3
.
According to
Smith et al.
(
2009
), Harmony is able to match larger schemas with
about 1,000 elements each. However, so far, no detailed evaluation of Harmony's
effectiveness and efficiency has been published.
5
Conclusions
We have provided an overview of selected approaches and current implementations
for large-scale schema and ontology matching. Commercial systems increasingly
support automatic matching but still have to improve much to better handle large
schemas. The current research prototypes share many similarities, particularly a
