An object knowledge canonical form /

Bibliographic Details
Main Author: Huarng, Kunhuang, 1961-
Other Authors: Childs, S. Bart (degree committee member.), Ellis, Newton C. (degree committee member.), Lively, William M. (degree committee member.)
Format: Thesis Book
Language:English
Published: 1993.
Subjects:
Online Access:ProQuest, Abstract
Link to OAKTrust copy
Description
Abstract:An Object Knowledge Canonical Form (OKCF) is introduced to facilitate object knowledge reuse in commercial object-oriented expert system building tools (ESBTs). The OKCF contains a meta class of facts and a meta structured fact to represent and translate object knowledge, including classes of facts and their instantiation (structured facts). For ease of understanding and maintenance, the OKCF is defined in Backus Naur Form (BNF). The OKCF is used to represent object knowledge. Three commercial ESBTs: CLIPS 5.0, KAPPA, and KEE are studied. The common and unique characteristics among these ESBTs are identified. Based on the OKCF, BNF rules for each of these ESBTs are proposed. These BNF rules include appropriate delimiters for each of these ESBTs and are demonstrated to represent object knowledge in these ESBTs. The OKCF also serves as an intermediate form in knowledge translation. The knowledge bases in source ESBTs are translated to the OKCF, then to target ESBTs. However, only the common characteristics among these ESBTs can be translated. Knowbase, a cross compiler, was built by the BNF rules to translate object knowledge between CLIPS 5.0 and KAPPA. The OKCF also provides properties for uncertainty management. The OKCF is used for uncertainty management systems (UMSs) in object-oriented ESBTs to resolve inheritance conflicts when uncertainties are involved. Besides, the OKCF is applied to translate similar and different UMSs among commercial ESBTs. Recommendations for future research include extension of the OKCF, translation of methods, and semantic interpretation to resolve inheritance conflicts.
Item Description:Vita.
"Major subject: Computer Science."
Physical Description:xiii, 206 leaves : illustrations ; 28 cm
Bibliography:Includes bibliographical references.