The authors give an overview of their blackboard-based, multiparadigm expert system development environment, euphemistically called BEST. According to this paper, their system is efficient and extremely rich in functionality. In fact, all the paradigms and techniques I ever heard of are embedded in BEST. It includes hybrid rule-based, object-oriented, procedural, and logic programming, with a combination of forward and backward chaining guided by an adaptive control and search strategy with efficient indexing (improved RETE algorithm). It provides model-based, temporal, nonmonotonic, hypothetical, explanatory, fuzzy, and approximate (with dynamic weighting) reasoning; opportunistic control; and automatic partitioning and scheduling. The blackboard architecture is used to integrate these paradigms, because of its inherent modularity. Knowledge is represented by frames in Prolog/REX. Control and strategy are expressed by meta-rules, so they can be enhanced. Systems created by this toolkit use a linear algebra package, spreadsheets, and database management systems.
From a formal point of view, the paper is excellent. It is not surprising that nothing from the theory could be explained in detail; differences from other implementations are only stated, but solutions are not shown. There are no concrete examples to illustrate the cooperation of the different knowledge sources. Two applications are mentioned, and one of them, an Investment Advisory Expert System, is introduced briefly. Unfortunately, half of this description deals with the connection of Excel through DDE; the authors do not explain the role and integrated work of the different paradigms implemented in this system. So the paper describes the authors’ impressive work but, as differences from other work are not explained in enough detail, the paper is not as interesting as it could have been.