Browse Prior Art Database

Design of a Composite Production Rule Syntax And Associated Logical Data Structure

IP.com Disclosure Number: IPCOM000102586D
Original Publication Date: 1990-Dec-01
Included in the Prior Art Database: 2005-Mar-17
Document File: 8 page(s) / 332K

Publishing Venue

IBM

Related People

Carlis, JV: AUTHOR [+2]

Abstract

As the understanding and use of expert system technology matures, it will be necessary to support larger and larger integrated knowledge bases. Supporting these bases will require many of the functions provided by traditional database management systems (such as data integrity, concurrent access, and indexes to increase performance). This project illustrates that it is possible to combine the syntaxes from several production rule systems into a composite syntax which encompasses all of the original rule systems. The composite can be used to integrate knowledge bases developed using original production rule systems, or it may be used directly in producing new knowledge bases.

This text was extracted from an ASCII text file.
This is the abbreviated version, containing approximately 18% of the total text.

Design of a Composite Production Rule Syntax And Associated Logical Data Structure

       As the understanding and use of expert system technology
matures, it will be necessary to support larger and larger integrated
knowledge bases.  Supporting these bases will require many of the
functions provided by traditional database management systems (such
as data integrity, concurrent access, and indexes to increase
performance). This project illustrates that it is possible to combine
the syntaxes from several production rule systems into a composite
syntax which encompasses all of the original rule systems.  The
composite can be used to integrate knowledge bases developed using
original production rule systems, or it may be used directly in
producing new knowledge bases. The composite is formed by identifying
the features that are supported by each rule system, and decomposing
the rules into the basic, common elements which are used to provide
each feature.  The process is analogous to determining the basic
elements (attributes) when designing an integrated application
database.  A composite syntax can then be built from the basic
elements and expressed in Backus-Naur Form (BNF).  The composite
syntax BNF describes how the basic elements are related to build
production rules and is used as a basis in doing a logical data
design for the rule data. The data design process results in a
conceptual data model of the data contained in the production rules
(see the figure).  This data model is analyzed to locate ambiguities
in, and simplify the BNF for, the composite syntax.

      There are a large number of systems available that use some
form of production rules to represent knowledge.  This project
examined systems that ranged from traditional approaches to
production rules, as well as newer approaches, such as computational
networks.  The systems examined included:
   -  Expert System Environment (ESE)
   -  KnowledgeTool*
   -  Intellicorp's Knowledge Engineering Environment (KEE**)
   -  Official Production System Version 5 (OPS5)
   -  A Generalized Network-Based Expert System Shell (AGNESS)
   -  SYLLOG

      Each system has its own rule syntax and a set of features that
it supports.  Features include:
   -  the types of information that can be referenced in the rules,
and
   -  the types of manipulation that can be done in the rules.

      A feature represents information or function that is supported
by the rule system.  The features determine the types of knowledge
that can be represented and how that knowledge can be used within
each system.  Some features are found in several or all systems, and
some are unique to a particular system.  Following is a summary of
the features that were found by analyzing the selected rule system
syntaxes.

      GLOBAL VARIABLES.  Data stored in a global variable is
available to all the rules.  The data is not copied for each rule,
rather the same s...