The University of Edinburgh -
Division of Informatics
Forrest Hill & 80 South Bridge

MSc Thesis #93115

Title:A Comparison of Simple Rule and Task-Based Diagnostic Systems in the Agroforestry Domain
Date: 1993
Abstract:We compare two different diagnostic systems: a simple rule-based and a task-based system in the agroforestry domain. The domain knowledge employed was drawn from the Indigenous Ecological Knowledge research programme, one of the research grants of the Department of Artificial Intelligence of the University of Edinburgh. Both systems were developed using comparatively simple approaches appropriate to the domain. The first prototype, the rule-based system, is a variant of a conventional backward chaining method. The second prototype, the task-based system, relies on a series of modular tasks that permit the control of the consultation. The results show the rigidness of the former against the flexibility of the latter with regard to the use of different diagnostic strategies. The flexible architecture of the task-based system permits not only inclusion of new diagnostic strategies but also implementation of the ability to choose the best strategies to apply. This is achieved by explicit and separate representation of the control knowledge from the domain knowledge of the task-based system, which is not feasible in the rule-based system because it integrates of both types of knowledge.

[Search These Pages] [DAI Home Page] [Comment]