Databases Reference
In-Depth Information
a
b
Fig. 9.7
Basic scenario expansion and synthetic scenario generation
create synthetic mapping scenarios that involve complex transformations coming
from a combination of transformations that the basic mapping scenarios describe.
For the generation of the instance of the source schema, STBenchmark generates a
ToXGene [
Barbosa et al. 2002
] configuration template with which one can invoke
ToXGene to produce the data of the source instance.
In the area of schema matching, the ISLab Instance Matching Benchmark
[
Ferrara et al. 2008
] is also following a bottom-up approach. It uses several algo-
rithms to create different data sets. It initially requires the creation of a reference
ontology for a specific domain. Then, this ontology is populated with instances by
querying Web sites. For example,
IMDB
enables the population of a
movie
ontology.
Subsequently, a number of modifications on the data takes place, with three goals
in mind: (1) to introduce variations in the data values, e.g., typographical errors,
(2) to introduce structural heterogeneity, e.g., properties represented by different
structural levels, aggregations, and others, and (3) to introduce local heterogeneity,
which mainly includes semantic variations that requires ontological reasoning to
cope with. Once the modifications have been performed, the benchmark users are
provided with the initial reference ontology and the modified one, against which
they evaluate matching tools.
6
Measuring Efficiency
6.1
Matching/Mapping Generation Time
Since one of the goals of mapping tools is to assist the matching/mapping designer
in performing the time-consuming matching and mapping tasks faster, time plays a
major role in measuring the performance of matching/mapping tools. Nevertheless,
mapping tools like Spicy [
Bonifati et al. 2008b
], HePToX [
Bonifati et al. 2005
], or
Clio [
Popa et al. 2002
], in their evaluation experiments, make only a small reference
