An Empirically-Derived Question Taxonomy for Task-Oriented Tutorial Dialogue
Kristy Elizabeth Boyer, William Lahti, Robert Daniel Phillips, D Michael, Mladen Alan Vouk, James C. Lester · 2009
Devising an expressive question taxonomy is a central problem in question generation. Through examination of a corpus of human-human task- oriented tutoring, we have found that existing question taxonomies do not capture all of the tutorial questions present in this form of tutoring. We propose a hierarchical question classification scheme for tutorial questions in which the top level corresponds to the tutor's goal and the second level corresponds to the question type. The application of this hierarchical classification scheme to a corpus of keyboard-to-keyboard tutoring of introductory computer science yielded high inter-rater reliability, suggesting that such a scheme is appropriate for classifying tutor questions in design-oriented tutoring. We discuss numerous open issues that are highlighted by the current analysis.