Databases Reference
In-Depth Information
mapping of heteroge neous information into the well-structured data model is not
generally easy.
Figure 5: Relation-based well-strucuted data model view
(3) Hybrid data models
The third approach is the
hybrid
approach. In this approach, data modeling
frameworks fit for different information sources are combined. For example, ADTs
are often incorporated into relational frameworks to accept semistructured data
such as XML documents (Figure 6). An advantage is that development of wrappers
is relatively easy. MIROWeb
10)
introduces the semistructured data types to accept
XML documents and XML-QL
49)
queries. Another possible advantage is that we
can exploit various functions provided by ADT methods. InfoWeaver
26,
35,
36)
features
dynamic conversion between relational data and the document type, and allows
users to apply ADT methods to the document data. The hybrid model can be
constructed based on the object model rather than the relational model. In this
case, ADTs correspond to object types in the object model.
Figure 6: Relation-and-ADT-based hybrid data model view
When we integrate data in different information sources, the system has to resolve
their structural heterogeneity. Facilities for this process heavily depend on
characteristics of the common data model. Well-structured data models usually
require that the heterogeneity be completely resolved when data is mapped into
the common data model. For example, WebLog's users have to decide how to map
Web information to the simple data structure it can deal with. Araneous provides
a language for mapping hypertext structures into relations. On the other hand, in
systems based on semistructured and hybrid data models, this resolution is
“postponed” until the user describes the integration specification. In this case,
languages for the integration specification often allow
regular path expressions
to
cope with structural heterogeneity.
3.2
Data Layout Specification
As mentioned in Section 1, data layout is also essential as well as the data
acquisition. A well-known example of layout specification languages is the HTML-
based template language for Strudel
19)
. Also, existing Web page authoring tools
can be used for the data layout specification
6)
. Authoring tools such as FrontPage
33)
