Databases Reference
In-Depth Information
Schema Comparison
This step is critical during schema integration. It provides the essential
knowledge used to merge two different schemas. The main task of this step
is to compare two distinct schemas and decide for each pair of compared
objects whether they represent the same real-world object or not. The result
of this step is a set of correspondence assertions that state which pairs of con-
cepts are semantically equivalent. Schema comparison is either done manu-
ally or supported by an automated tool. In the first case, equivalence
assertions are elaborated by the designer [28]. In the second case, they are
discovered by a structure-matching algorithm [29], although the validation
of the assertions is done by a human designer, due to the high level semantics
required to enforce the equivalence between concepts.
To illustrate some of the problems that can arise during the comparison
step, let us consider the two E/R schemas in Figure 13.13. The comparison
of entities must state that Worker and Employee, in schema 1 and schema 2,
respectively, are semantically equivalent despite the fact that they have
different names and different attributes, while Worker and Department
are not semantically equivalent despite the fact that they have identical
attributes.
If there are only a few small schemas, the comparison can be done
manually. The correspondence assertions established by the human designer
are generally of high semantic quality; they can be used profitably as inputs
to the remaining integration steps. However, in large application domains,
the comparison process becomes tedious and complex as a human activity.
Different comparison tools have been proposed to aid in this combinatory
task [29]. However, the produced equivalence assertions are of low-level
semantics, and a human intervention is again necessary to elicit those that
are pertinent. For example, syntactic comparison of the schemas given in
Schema 1
Schema 2
Worker
Car
Employee
Soc-security-nb
Name
Date of birth
Address
Department
Name
Address
Name
Address
Car_number
Year
Owns
Owner
Object
Works-in
Worker
Place
Figure 13.13
Two initial schemas.
