Databases Reference
In-Depth Information
O
Source
Anchors
O
Target
C
1
B
S2
C
1
B
S1
C
2
C
4
C
5
C
6
C
2
C
3
C
3
C
7
C
8
C
11
C
12
C
6
C
5
C
4
C
9
C
10
C
13
C
9
B
S3
C
8
C
10
C
7
B
T1
B
T3
B
T2
Fig. 1.3
Partition-based matching in Falcon-AO and Taxomap (from
Hamdi et al.
(
2009
))
segments) are located around the anchors, and their size depends on the continued
success of finding match partners for the considered elements.
Zhong et al.
(
2009
) focus on the case when a small ontology is matched with a
much larger one, e.g., one that is obtained from merging several others. They deter-
mine the subontology (partition) from the larger ontology that is most similar to the
smaller ontology and consider only this subontology for matching to improve effi-
ciency. Finding the subontology is performed in two steps. First, a name matcher is
applied on the Cartesian product of elements to determine the most similar ontology
elements from the large ontology. Then, the subontology is determined by evaluating
the subgraphs around the similar elements found in the first step.
3.2
Parallel Matching
A relatively straight-forward approach to reduce the execution time of large-scale
matching is to run match processes in parallel on several processors. As discussed
in
Gross et al.
(
2010
), two main kinds of parallel matching are applicable: inter-
and intra-matcher parallelization.
Inter-matcher parallelization
enables the paral-
lel execution of independently executable (parallel) matchers in match workflows.
This kind of parallelism is easy to support and can utilize multiple cores of a single
computing node or multiple nodes. However, inter-matcher parallelization is limited
by the number of independent matchers and not applicable for sequential matchers.
Furthermore, matchers of different complexity may have largely different execu-
tion times limiting the achievable speedup (the slowest matcher determines overall
execution time). Moreover, the memory requirements for matching are not reduced
since matchers evaluate the complete ontologies.
