<?xml version="1.0" standalone="yes"?>
<Paper uid="N03-2018">
  <Title>Towards Emotion Prediction in Spoken Tutoring Dialogues</Title>
  <Section position="2" start_page="0" end_page="0" type="intro">
    <SectionTitle>
1 Introduction
</SectionTitle>
    <Paragraph position="0"> Connections between learning and emotion are well-documented (Coles, 1999), and studies have shown considerable benefits of spoken tutoring (Hausmann and Chi, 2002). Human tutors can respond to both the content of student speech and the manner with which it is spoken (e.g. 'confidently' or 'uncertainly'), but most intelligent tutoring dialogue systems are text-based and thus limited in their ability to recognize such learning states (Rose and Freedman, 2000; Rose and Aleven, 2002). Building spoken dialogue tutoring systems has great potential benefit, for speech is the most natural and easy to use form of natural language interaction, and it supplies a rich source of prosodic and acoustic information about the speaker's current mental state, which can be used to monitor the pedagogical effectiveness of student-computer interactions. The success of computer-based tutoring systems could increase if they predicted and adapted to student emotional states, e.g. reinforcing positive states, while rectifying negative states (Evens, 2002).</Paragraph>
    <Paragraph position="1"> Although (Ang et al., 2002; Litman et al., 2001; Batliner et al., 2000) have hand-labeled naturally-occurring utterances in a variety of corpora for various emotions, then extracted acoustic, prosodic and lexical features and used machine-learning techniques to develop predictive models, little work to date has addressed emotion detection in computer-based educational settings. In this paper we describe preliminary annotation of positive, negative, and neutral emotions in a human-human tutoring corpus and discuss the results of pilot machine learning experiments whose goal is to develop computational models of specific emotional states (Section 3) for use in a spoken dialogue system (Section 2).</Paragraph>
  </Section>
class="xml-element"></Paper>