HAL: a faster match algorithm

Abstract

Existing match algorithms treat the matching process like the querying process of relational databases. Owing to the combinatorial nature of the matching process, the match time greatly varies in different recognize-act cycles. Current match algorithms utilize local matching support networks with redundant working memory elements shared among rules involving the same classes. Since the match time is a dominant factor in the total execution time of a production system, such large match time makes production systems with existing match algorithms unsuitable for many applications. To reduce match time, we introduce the Heuristically-Annotated-Linkage (HAL) match algorithm. HAL differs from traditional match algorithms in that HAL employs a fixed-traversal-distance pseudobipartite network approach of treating rules and classes as objects, or nodes, in only one global pseudobipartite-graph-like connection and communication scheme. In addition, HAL is more efficient than other existing match algorithms because it is capable of immediate characterization of any new datum upon arrival. This paper reviews existing match algorithms, presents HAL, and analyzes the performance of HAL in comparison with existing algorithms.