Getting Ready for Session #13 (October 9, 1996)
Knowledge & Knowledge Representation: Search to Support Inference
Durkin, J. (1994). Expert Systems: Design and Development. Englewood Cliffs, NJ: Prentice-Hall. [QA76.76.E95 D87 1994b] & [PAM] pp. 100-127
Knowledge representation influences search time as well as the possibility of a successful search. For instance, if some attribute of an object is not represented (e.g., type of research report), the searcher might need to scan each item rather than be informed of those that match some criteria (e.g., clinical trial). Search spaces for knowledge bases tend to "mushroom," quickly eliminating the possibility of finding a solution within an acceptable time period. The trial and error search strategies that are typically employed do not take advantage of possible knowledge structures that could be provided as part of a knowledge base.
This selection introduces such ideas and approaches as forward and backward chaining, conflict resolution, goals, demons, and the general search strategies that are employed to find something in a problem space. So, forward and backward chaining along with the mixture of the two are processes or strategies for specifying a search space. Search strategies, such as breadth first or depth first are search approaches that operate on the search space specified by forward or backward chaining.