The Improvement for Rete Algorithm

Rete algorithm is the most efficient forward inference algorithm, widely used in rule engine systems. The Rete algorithm can be divided into two parts: rule compilation and runtime execution. Rule compilation describes how the rules in the production memory generate an efficient discrimination network - Rete network, while runtime execution describes how the asserted facts propagate through the network and fire the rule execution. In the rule compilation part, this paper introduces ´or´ connective between conditions, then proposes an algorithm that interprets a single production containing multiple ´or´ connectives as the equivalent of multiple productions. In runtime execution part, a specific rule engine handles common objects as facts to provide easy integration, and this paper introduces the ´update´ primitive to update facts in the working memory of the rule engine and describes how to implement it through a shadow proxy mechanism.