Specifying Meta-Level Architectures for Rule-Based Systems

Abstract

Explicit and declarative representation of control knowledge and well-structured knowledge bases are crucial requirements for efficient development and maintenance of rule-based systems. The CATWEAZLE rule interpreter allows knowledge engineers to meet these requirements by partitioning rule bases and specifying meta-level architectures for control. Among others the following problems arise when providing tools for specifying meta-level architectures for control: 1. What is a suitable language to specify meta-level architectures for control? 2. How can a general and declarative language for meta-level architectures be efficiently interpreted? The thesis outlines solutions to both research questions provided by the CATWEAZLE rule interpreter: 1. CATWEAZLE provides a small set of concepts based on a separation of control knowledge in control strategies and control tactics and a further categorization of control strategies. 2. For rule-based systems it is efficient to extend the RETE algorithm such that control knowledge can be processed, too.