Semantic networks have been used to represent knowledge in a variety of
problem domains. A semantic network can graphically model the concepts
and complex relationships that exist in a real world application.
However this complexity also makes it difficult to maintain the semantic
network over time as the model must be modified to reflect the dynamic
changes present in most application areas. For example, the meaning
of certain terms and relations can shift over time in a dynamic performance
environment. This need for the constant adaptation of the system
to an ever-changing environment led us to explore the use of evolutionary
computational techniques as a tool for re-engineering semantic networks.
The manual re-engineering of a knowledge base will be extremely time-consuming
and difficult to accomplish due to the complexity of the knowledge structure.
This complexity includes both the dimensionality and decomposability of
the semantic network. In this project we will utilize the evolutionary
computational process, known as Cultural Algorithms, to learn to
re-engineer semantic networks in order to improve their performance in
a large-scale manufacturing application.