Extending Loosely Coupled Federated Information Systems Using Agent Technology - Distributed Artificial Intelligence, Agent Technology, and Collaborative Applications - page 119

hide the heterogeneity of the component data sources in the federation. Uniform query languages such
as MDBQL (multidatabase query language) have evolved to create the requisite binding between the
presentation layer and the federation layer (Busse et al.). MDBQL can be used to create specific cus-
tom-defined global views from the participating schema in order to reduce the overhead on the users
trying to interact with the foundation layer. With the increasing prominence of Web services and service
composition models for quick and easy information interchange, loosely coupled federations are evolving
as the sensible alternative compared to tightly coupled federations (Zimmermann, Krogdahl, & Gee,
2004). However, the dynamics of the Internet environment poses formidable integration challenges,
especially if the FIS is to allow organizations to join or leave the federation with a certain desired level
of trust, ease, freedom, and security.
The importance of generating a correspondence between the logical schema and federated layer
for conducting queries in an FIS is very evident (Busse et al., 1999). One meaningful way to express
correspondence is through the use of ontological descriptions defined by humans or semiautomatically
generated by computer systems using newer standards of ontology description languages like OWL,
OWL-S (OWL for semantic Web services), or BPEL (business process execution language; OWL-S ,
n.d.). Grounded in robust theoretical foundations based on description logic (DL; Baader, Calvanese,
McGuinness, Nardi, & Patel-Schneider, 2003), OWL has evolved as the W3C (World Wide Web Con-
sortium) standard for semantic knowledge representation (Thomas et al., 2005). In conjunction with
XML (extensible markup language), which is widely supported in almost all database management
systems, interoperability challenges between schema definitions are significantly lowered since DL-
based XML schema can be generated with relative ease ( OWL-S ). The inclusion of semantic support
in common query languages also offers software applications and agents the ability to execute KQML
(knowledge query and manipulation language) performatives with a strong semantic focus (Labrou &
Finin, 1997). Many commercial tools are now available on the market (Protégé, http://protege.stanford.
edu; Racer, http://www.racer-systems.com) that can generate domain ontology from DL and verify its
consistency and conformance against schema models. OWL-S-based ontology documents can store
all service properties of a data source and advertise service capabilities that can be used for automatic
service discovery and composition in alignment with the fundamental concepts of service composition
architectures (SCA, 2005).
The growing popularity of semantic Web models has drawn a stronger impetus toward the use of
intelligent agents to discover, invoke, compose, and monitor services using ontology documents in an
automated or semiautomated way (Thomas et al., 2005; Zou et al., 2003). Agent technology has evolved
as critical middleware in a variety of distributed systems, both peer to peer as well as client-server, and in
numerous other software component architectures that use the agent paradigm (FIPA, n.d.; JADE Project,
n.d.). The FIPA agent interoperability specifications are designed to facilitate end-to-end internetwork-
ing of intelligent agents and lays down standards for a variety of implementation-specific stipulations
such as agent communication languages, the message transport protocol for IIOP, quality of service,
and ontology specifications (FIPA; Open Financial Exchange [OFX], 2006). Almost all popular agent
development environments (such as MadKit, JADE, etc.) allow extensive and flexible agent configura-
tion and development capabilities and are compliant with the FIPA interoperability standards.
In an FIS, the objective of semantic integration is undertaken at the federation layer. The different
steps in the semantic integration process include collection, fusion, abstraction, and supplementation
(Busse et al., 1999). In a tightly coupled FIS, this is rigid and fixed since changes to local data sources
are least probable or not allowed. Given the dynamic environment of the Internet, a tightly coupled