Circularity in Rule Knowledge Bases Detection using Decision Unit Approach

摘要

This paper presents the usage of decision units in circularity detection task. Decision units allow modular organisation of the rule knowledge base, which facilitate programming and base verification, simultaneously increasing the clarity of achieved results and ergonomics of a knowledge engineer labour. The decision units are simple and intuitive models describing relations in rule knowledge base, being direct superstructure of a rule-base model. Decision units allow us to reduce search space in circularity detection task. Thus, the decision units can be considered as a source of simple heuristics, which make verification simple and more efficient. In this same way we can reduce search spaces in other verification tasks, i.e. in redundancy detection in rules chains. We can take into account only selected paths in rules chains and performing verification more quickly. The knowledge base shown as a decision unit net allows us to make a convenient presentation in the graphic form - both on a computer screen and on the prints. Thus, we can present verification result in clear and user friendly form. The difficulties with rule base verification have been described in a natural and intuitive way. Detected anomalies and suggested methods of their elimination can be presented to the knowledge engineer in a suggestive way, without the necessity of relating to a complex, conceptual issues.