Information Technology Reference
In-Depth Information
loosely coupled FISs are becoming the preferred mode for information exchange over the Web, the ben-
efits of languages based on open standards for semantic markup such as OWL-S and DAML (DARPA
[Defense Advanced Research Projects Agency] Agent Markup Language) cannot be ignored (MadKit
Project, n.d.; Martin, Burstein, Lassila, Paolucci, Payne, & McIlraith, 2003;
OWL-S
, n.d.). In addition,
the importance of data service discovery over the Web (using UDDI and WSDL [Web service definition
language]) for identifying and dynamically retrieving relevant information from different content sources
is no longer the result of a simple predetermined query. Successful information retrieval is thus a direct
function of the clarity and precision of the semantics of the request: what we really want from a query.
As flexibility to allow data sources to come and go in an FIS becomes indispensable, it is imperative
that new mechanisms are developed that will allow extensible agent (new or existing) architectures to
exploit all available data resources when returning information that is relevant to the user. Currently
available agent development environments support agent negotiations and task delegations, but lack
architectural templates that will allow agent creation and modification based on semantic matchmaking
between the data source and the defined capabilities of the agent.
A description of the role of each agent in the agent community and its purpose in the context of the
running example (from the previous sections) is given below.
Personal Agent
Personal agents (PAs) are the end-user interfaces that take the user context and the query clause. They
ensure that the requests placed by the user are valid and pass the query packaged in KQML to the
extended directory facilitator (EDF). All agents that interact with the user are grouped into an agent
community that resides at the presentation layer of the FIS. The EDF determines the relevant data
source on which the request can be processed. Appropriate interaction controller (IC) agents are also
identified based on agent capability (ask all, ask specific), agent functionalities (O-O [object-oriented]
query processing), restrictions imposed, supported ontologies (finance), ontological language capaci-
ties (BPEL, OWL-S), and adaptivity (able to be cloned, replicable; Nodine et al., 1999). The PA acts as
a gateway between the EDF and the end user with the authority to query, update, or modify data from
the data sources in an FIS.
Table 1. Sequence of communication and OWL-S profile descriptions for the multiagent architecture
A user in FIS provides query specifications
Example:
Loan Oficer
•
Loan Officer needs information about employment history, financial
stability, credit check and current property value estimate.
•
Invokes service using Personal Agent.
Personal Agent
PA returns results to the loan officer.
continued on following page
Search WWH ::
Custom Search