Towards an Agent-Based Knowledge Management System - Developing Concepts in Applied Intelligence - page 96

(KBS) of the OM; (2) Reasoning Agent (RA) that will be involved with CBR tasks;
(3) Learning Agent (LA) that will be responsible for modifying the OM and Task
Agent (TA) that will be responsible for identifying when and how to apply KC. TA
will be responsible for identifying to what level a specific problem belongs, whether
in the CBR-Base or the KB. RA will investigate the CBR-base to identify similar
cases to the given problem, and then RA will find percentages of similarities to cases
and give the results back to the TA. If a desired case is founded then no need to call
KCA. Otherwise, TA will call KCA in order to find the appropriate experts who may
give a consultation about the problem.
The different components of the system are illustrated in Figure 1 below.
Fig. 1. A Learning KMS.
OM is divided into two parts: one for KBs of related entities and the other for
CBR-Bases. Figure 2 illustrates the structure of an OM. There are three knowledge
levels. An ontology is employed to structure the case bases into hierarchies (cf. Figure
3). It stores an abstraction and the mapping of all domain-specific concepts and mean-
ings and can be used to verify the existence of all acquired concepts in KB. A similar-
ity measure is defined in order to retrieve similar cases. For each attribute of a given
case and a given query, a similarity can be calculated. The attribute similarities can be
combined to an overall similarity value. The employment of similarity measure makes
it possible to rank the solutions.ïïA case represents a previous experience of a situa-
tion. It consists of two parts: a problem description together with the con-
text/environment in which it has occurred and a solution to this case. KB can be ac-
cessed during the retrieve phase using a similarity based measure. When presented
with a particular case, DM calls TA in order to identify the type/level of problem
based on its parameters. Weights can be associated with parameters to indicate their
importance. Weights help in fine-tuning and reaching more accurate decisions during
reasoning by enabling RA to retrieve the most similar and relevant cases. TA will
generate a unique problem-ID and store the problem in the problem-base. RA can use
the problem-ID and ontology in order to investigate the level to which the problem