Analyzing Dependencies Between Student Certainness States and Tutor Responses in a Spoken Dialogue Corpus

Abstract

This study presents an approach for developing more empirically motivated affective dialogue tutorial systems. In particular, we use n-gram techniques from statistical natural language processing to identify dependencies between student affective states of certainty and subsequent tutor dialogue acts, in an annotated corpus of human–human spoken tutoring dialogues. We first represent our dialogues as bigrams of annotated student and tutor turns. We next use χ² analysis to identify dependent bigrams, i.e., where the student certainty and tutor dialogue act annotations are related in some way other than predicted by chance. Our results show dependencies between many student states and subsequent tutor dialogue acts. We then analyse the dependent bigrams both with respect to differences between observed and expected counts and with respect to correlations with learning; these analyses suggest ways that our current computer tutor can be enhanced to adapt its dialogue act generation based on these dependencies.