Databases Reference
In-Depth Information
Tabl e 9. 2
Element s
tructuralities for the intended schema of Fig.
9.10
Element
P
int
P
gen
Element structurality
max
.0;
2
1
.1
1/
2
B
A
A
/
D
1
1
max
.0;
2
1
.1
1/
21
D
A
A
/
D
1
max
.0;
2
2
.2
2/
2
E
A,D
A,D
/
D
1
2
max
.0;
2
0
.0
0/
22
G
A,D
;
/
D
0
max
.0;
2
1
.2
1/
2
1
4
C
A,B
A,D
/
D
2
max
.0;
2
1
.1
1/
22
1
2
F
A,D
A
/
D
There has been an interesting set of experimental results [
Duchateau 2009
]on
computing the above metrics using a number of different datasets with the two
popular matching systems: COMA
[
Aumueller et al. 2005
] and Similarity
Flooding [
Melnik et al. 2002
]. The former system builds integrated schemas using
an ASCII-tree format (then converted into XSD using a script [
Duchateau et al.
2007
]), while the latter system directly generates an XSD integrated schema. The
matches discovered by the tools before building the integrated schema have not been
checked. The experiments include a dataset extracted from the XCBL
7
and OAGI
8
collections, a dataset on university courses provided by the Thalia benchmark [
Ham-
mer et al.
2005
], a Biology dataset from Uniprot
9
and GeneCards,
10
a currency and
sms dataset,
11
and a university department dataset [
Duchateau et al. 2008
]. These
datasets present various features that reflect real-world scenarios. For instance, the
biology dataset contains a specific vocabulary that is not usually found in common
dictionaries. The dataset about university courses describes a case in which many
schemas have to be integrated. A part of the experimental results obtained from that
effort is illustrated in Fig.
9.11
. It has been noticed that the tools can obtain a high
completeness in most cases, mainly because the tools promote precision during the
matching phase. On the contrary, the minimality is more difficult to achieve, since
it depends on the recall. Finally, structurality is mostly preserved because the tools
try to keep the same structure that they find in the source schemas.
CC
8Conclu ion
We have presented a retrospective on key contributions in the area of evaluating
matching and mapping tools. Schema matching and mapping is a relatively new
area that has received considerable attention in the last few years. Since these
7
www.xcbl.org.
8
www.oagi.org.
9
http://www.ebi.uniprot.org/support/docs/uniprot.xsd.
10
http://www.geneontology.org/GO.downloads.ontology.shtml.
11
www.seekda.com.
