SCoT: a spoken conversational tutor

  • Authors:

  • Karl Schultz;Brady Clark;Heather Pon-Barry;Elizabeth Owen Bratt;Stanley Peters

  • Affiliations:

  • Center for the Study of Language and Information, Stanford University, Stanford, California;Center for the Study of Language and Information, Stanford University, Stanford, California;Center for the Study of Language and Information, Stanford University, Stanford, California;Center for the Study of Language and Information, Stanford University, Stanford, California;Center for the Study of Language and Information, Stanford University, Stanford, California

  • Venue:
  • AAAI'04 Proceedings of the 19th national conference on Artifical intelligence
  • Year:
  • 2004

Quantified Score

Hi-index 0.00

Visualization

Abstract

We describe SCoT, a Spoken Conversational Tutor, which has been implemented in order to investigate the advantages of natural language in tutoring, especially spoken language. SCoT uses a generic architecture for conversational intelligence which has capabilities such as turn management and coordination of multi-modal input and output. SCoT also includes a set of domain independent tutorial recipes, a domain specific production-rule knowledge base, and many natural language components including a bi-directional grammar, a speech recognizer, and a text-to-speech synthesizer. SCoT leads a reflective tutorial discussion based on the details of a problem solving session with a real-time Navy shipboard damage control simulator. The tutor attempts to identify and remediate gaps in the student's understanding of damage control doctrine by decomposing its tutorial goals into dialogue acts, which are then acted on by the dialogue manager to facilitate the conversation.