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.