Speech and Affect in Intelligent Tutoring Dialogue Systems

Dr. Diane Litman
LRDC & Computer Science Department

While human tutors typically interact with students using spoken
dialogue, most computer dialogue tutors are text-based.  To evaluate
the utility of adding spoken language capabilities to dialogue tutors,
we have developed ITSPOKE (Intelligent Tutoring SPOKEn Dialogue
System), a speech-based dialogue system that uses the Why2-Atlas
qualitative physics tutoring system (VanLehn et al.) as its
``back-end.''

In this talk, I will first describe ITSPOKE.  I will then present
results from two empirical studies of corpora of human-human and
human-computer spoken dialogues.  The first study addresses how
speech-based dialogue tutoring impacts student learning, while the
second address how student attitudes and emotions can be automatically
detected from features of student utterances.