Using Discourse Structure to Model Student Performance in Speech-based Tutoring Systems

Mihai Rotaru, PhD student, Pitt CS

Tuesday, November 28
Noon - SENSQ 5317
Free pizza for attendees starting at 11:45 a.m.

Hosted by Diane Litman

Abstract

Human one-on-one tutoring has been shown to be the most effective form of instruction. Intelligent tutoring systems are computer systems that attempt to replicate the effectiveness of human tutors. Not surprisingly, there is still a big performance gap between these systems and the human tutors. To bridge this gap, it is important to understand what are the factors that influence or contribute to student's learning while interacting with such systems. Previous work has looked at various factors (e.g. correctness) but, in general, these factors are measured across the entire interaction (e.g. how many times the student was correct overall). In this work, we propose taking into account the context in which the performance factors are being measured. To define the context, we exploit the hierarchical aspect of the discourse structure. We validate our proposed approach by testing two hypotheses. First, we hypothesize that correctness at specific places in the interaction is more informative for performance modeling than overall correctness. Second, we hypothesize that the dialogues with students that learn more exhibit other patterns than the dialogues with students that learn less. Our empirical analysis confirms the two hypotheses, highlighting the importance of discourse structure for modeling student performance in intelligent tutoring systems.

Biography of Speaker

Mihai Rotaru is a 6th year PhD student in the Department of Computer Science, University of Pittsburgh, USA. He works under the supervision of Dr. Diane J. Litman. He was born in Romania and received his B.Sc. and M.Sc. from West University, Timisoara, Romania. He is primarily interested in Spoken Dialogue Systems with a focus on Intelligent Tutoring Systems. He is also interested in several other related areas: Natural Language Processing, Machine Learning and Artificial Intelligence. His non-academic interests include good movies, good books, traveling, exercising, cool gadgets and the company of his friends.