Most current approaches to conversation system development invoke a complex set of language parsing rules or development tools difficult for novices to handle and are unable to convincingly simulate advanced natural language features such as topic awareness or conversation thread involvement. This study proposes an alternate approach based on the ripple down rules (RDR) algorithm, presently used to enable more effective maintenance of expert systems. This tree-based algorithm enables a conversation system to travel incrementally deeper into a particular topic, then to switch based on context-dependent information to the correct previously discussed topic � resulting in a highly reusable method of developing conversation systems based around a variety of topics.
History
Publication title
Journal of Advanced Computational Intelligence
Volume
14
Issue
6
Pagination
741-745
ISSN
1343-0130
Department/School
School of Information and Communication Technology
Publisher
Fuji Technology Press
Place of publication
Japan
Rights statement
Copyright (C) 2010 Fuji Technology Press Co,.
Repository Status
Open
Socio-economic Objectives
Information systems, technologies and services not elsewhere classified