Information Technology Reference
In-Depth Information
By the way, a nationwide OPAC system can also integrate all library databases in a
country. This integration is achieved by the technique that all databases in an OPAC
system must follow the same specification, that is, an OPAC system consists of
homogeneous (or same structure) databases. The remarkable innovation of an RSS,
different from OPAC systems, is that an RSS can integrate heterogeneous (or
different structure) databases on the Internet. This is crucial to area studies, because
researchers use and need heterogeneous databases (e.g., libraries, archives, museums).
For example, the NIHU RSS [10] seamlessly integrates more than 100 heterogeneous
databases that five institutes hold. The CIAS also opens its RSS for area studies [11].
An RSS is modeled as 3-layered structure shown in Figure 3. The first layer is
“Database Layer” where original databases are allocated. Databases in this layer are
required to encode data by XML. An RRS needs frequent data conversion and
exchange processes between databases. Since data structure is different from each
database and data definitions are often modified, data conversion process must be
easy to follow these differences and modifications uniformly and independently from
databases. In addition, since binary data encoding and structures are different from
database systems, “data portability” is necessary while exchanging data between
database systems. XML has effective data conversion methods (XSLT) and data
portability function between heterogeneous information systems.
The second layer is “Metadata Layer”. Standard metadata is essential to “hub-data”
that gives an independent record structure from databases. In an RSS, records of each
database are converted into (or mapped onto) the hub-data records, by which all
records of databases integrated into an RSS have the same structure. Figuratively
speaking, Metadata Layer is an envelope that wraps databases, and users can see all
databases in this envelop as one database (i.e., one record structure). Selection or
definition of metadata decides the performance of RSSs.
The third layer is “Information Retrieval Layer”. Metadata specifications define
only record structures. Even if database systems integrated into an RSS adopt the
same metadata, retrieval procedures may be different from database systems.
Database systems in this layer are required to cope with a standard retrieval
procedure. Figuratively speaking, Information Retrieval Layer is also an envelope that
wraps database systems, and users can see all database systems within this envelop as
one database system (i.e., one record structure and one retrieval procedure).
Spatiotemporal attributes are often difficult to organize as metadata: there are many
kinds of spatiotemporal attributes in the world, but most metadata have few elements
for them. To solve this problem, some RSSs expand original metadata by adding their
own elements. The NIHU RSS is an example. Since the purpose of the NIHU RSS
was to integrate heterogeneous databases, a minimum element set was thought to be
the best metadata, and NIHU therefore adopted the Dublin Core Metadata Element
Set (DCMES) [12] as its base metadata. DCMES is simple and widely used on the
Web, but it is too small to describe precise spatiotemporal attributes. To describe a
variety of spatiotemporal attributes of humanities' resources, the NIHU RSS adds
spatiotemporal elements to expand the original DCMES [13].
On the other hand, the CIAS RSS is designed to organize wide range of resources
collected by researchers of area studies. Since these resources are collections of field
notes, photographs, maps, and so on, minimum elements' sets are not enough for the
purpose. The CIAS RSS needs larger metadata sets [13], and introduces combinations
