A New Buggy Rule and Template-Template Based Tutorial Dialogue System

Natural Language Generation in Spoken and Written Dialogue

Volume

Papers from the 2003 AAAI Spring Symposium

Contents

To simplify the time consuming authoring task of template generation, we have developed a new buggyrule and template-template based tutorial dialogue system which has not only simplified the authoring but has markedly improved the system performance. The template-template architecture comprises two schemes; extraction rules are used to expand the classical templates into a variety of template patterns by introducing rule-based symbols in some of transitional nodes of the template automata while the buggy rules introduced are capable of identifying and hence generating bugs from learners’ erroneous responses automatically. Many different templates can now be integrated into, or equivalently extracted from, a single template-template so that a considerable reduction is achieved in both space and time complexity of the overall system performance. The new template-template automaton is expected to play an important role in many other applications including intelligent tutoring systems of both text-based or speech-based dialogue systems, voice-enabling call center or voice portal systems or in fact any systems focused on enhanced as well as advanced human computer interfaces, implementing more natural human computer interactions between the system and humans.