Rule–based expert systems proved to be a successful AI technology in a number of areas. Building such systems requires creating a rulebase, as well as providing an effective inference mechanism that fires rules appropriate in a given context. The paper briefly discusses main rule inference algorithms Rete, TREAT and Gator. Since large rulebases often require identifying certain rule clusters, modern inference algorithms support inference rule groups. In the paper the case of the new version of Drools, introducing the RuleFlow module is presented. These solutions are contrasted with a custom rule representation method called XTT2. It introduces explicit structure in the rulebase based on decision tables linked in an inference network. In this case, the classic Rete–based solutions cannot be used. This is why custom inference algorithms are discussed. In the paper possible integration of the XTT2 approach with that of RuleFlow is discussed.
[1]
Charles L. Forgy,et al.
Rete: A Fast Algorithm for the Many Patterns/Many Objects Match Problem
,
1982,
Artif. Intell..
[2]
Grzegorz J. Nalepa,et al.
HeKatE Rule Runtime and Design Framework ?
,
2009
.
[3]
Ernest Friedman-Hill,et al.
Jess in action : rule-based systems in Java
,
2003
.
[4]
Grzegorz J. Nalepa,et al.
The HeKatE methodology. Hybrid engineering of intelligent systems
,
2010,
Int. J. Appl. Math. Comput. Sci..
[5]
Paul Browne,et al.
JBoss Drools Business Rules
,
2009
.
[6]
John Durkin,et al.
Expert Systems
,
1994
.
[7]
Daniel P. Miranker.
TREAT: a better match algorithm for AI production systems
,
1987,
AAAI 1987.