Information Technology Reference
In-Depth Information
ate decision rules for the problem domain. The repository contains information about the context and
syntactical representation of the information. This information provides the domain attributes pertinent
to the decision problem. This generates domain specific knowledge in the form of domain attribute in-
formation and rules for making decisions in the specific problem context. The system ensures that this
knowledge is generated in a format conducive for sharing and use of the information across a distributed
and heterogonous platform.
Knowledge Agents
We use the domain knowledge object as the modular abstraction for knowledge representation and
knowledge exchange facilitation in IKMDSA. Domain knowledge objects are made available to agents
by the learning agent sharing the object with the knowledge agent. The knowledge agent manages the
knowledge available in IKMDSA and allows for other agents in the system to know of, request and
receive the domain knowledge in the system. The system utilizes the domain knowledge object as the
modular knowledge abstraction for communication of knowledge across the multiple agents of the
system. Therefore, when the domain knowledge object is shared with an agent of the system, the agent
becomes aware of the problem context descriptions, in addition to the rules that govern decision-making
in the specific problem context. The knowledge agent is also responsible for the maintaining the col-
lection of domain knowledge available in the system through interactions with a knowledge repository.
The knowledge agent contains methods to generate rules to support ad-hoc queries by the user agent.
This is supported through the interactions of the knowledge agent with the knowledge repository of
the system that is implemented as a set of XML documents that can be stored in a repository that is
capable of storing XML documents such as the Oracle 9i family of information management products.
This knowledge repository allows for the easy storage and retrieval of the knowledge contained in a
domain knowledge object. Thus, the knowledge is available to all the agents in the system through the
activities of the KM behaviors of the knowledge agent object. In this respect, the interactions among
the agents in this system are modeled as collaborative interactions, where the agents in the multi-agent
community work together to provide decision support and knowledge-based explanations of the deci-
sion problem domain to the user.
As shown in Figure 2, users of IKMDSA interact with the system through user agents that are con-
stantly aware of all domain knowledge contexts available to the system, through a list of names of the
domain knowledge objects that is published and managed by the knowledge agent. This allows every
user agent, and hence every user, to be aware of the entire problem space covered by the system. The
user agent requests and receives the knowledge available for a specific problem domain by making a
request to the knowledge agent, at the behest of the user. The knowledge agent, upon receiving this
request, shares a domain knowledge object with the user agent, thereby making problem domain infor-
mation and decision rules available to the user agent. The knowledge agents also serve as the means
to service any ad-hoc queries that cannot be answered by the user interface agents, such as queries
regarding knowledge parameters that are not available to the user interface agents. In such cases, the
Knowledge agent, with direct access to the knowledge repository can provide such knowledge to the
user agents, for the benefit of the user. This information is shared in the form of two W3C compliant
XML document object model (DOM) objects, domain and rules, which represent an enumeration and
explanation of the domain attributes that are pertinent to the problem context and the rules for making
decisions in the specified problem context. Once the domain knowledge object is available to the user
Search WWH ::
Custom Search