<?xml version="1.0" standalone="yes"?>
<Paper uid="W03-0201">
  <Title>Marineau Heather Hite-Mitchell</Title>
  <Section position="3" start_page="0" end_page="0" type="intro">
    <SectionTitle>
2 AutoTutor
</SectionTitle>
    <Paragraph position="0"> AutoTutor is an ITS applicable to any content domain.</Paragraph>
    <Paragraph position="1"> Two distinct domain applications of AutoTutor are available on the Internet, for computer literacy and conceptual physics. The computer literacy AutoTutor, which has now been used in experimental evaluations by over 200 students, tutors students on core computer literacy topics covered in an introductory course, such as operating systems, the Internet, and hardware. The topics covered by the physics AutoTutor are grounded in basic Newtonian mechanics and are of a similar introductory nature. It has been well documented that AutoTutor promotes learning gains in both versions (Person et al. 2001).</Paragraph>
    <Paragraph position="2"> AutoTutor simulates the dialog patterns and pedagogical strategies of human tutors in a conversational interface that supports mixed-initiative dialog. AutoTutor's architecture is comprised of seven highly modular components: (1) an animated agent, (2) a curriculum script, (3) a speech act classifier, (4) latent semantic analysis (LSA), (5) a dialog move generator, (6) a Dialog Advancer Network, and (7) a question-answering tool (Graesser et al. 1998; Graesser et al. 2001; Graesser et al. 2001; Person et al. 2000; Person et al.</Paragraph>
    <Paragraph position="3"> 2001; Wiemer-Hastings et al. 1998).</Paragraph>
    <Paragraph position="4"> A tutoring session begins with a brief introduction from AutoTutor's three-dimensional animated agent.</Paragraph>
    <Paragraph position="5"> AutoTutor then asks the student a question from one of topics in the curriculum script. The curriculum script contains lesson-specific tutor-initiated dialog, including important concepts, questions, cases, and problems (Graesser and Person 1994; Graesser et al. 1995; McArthur et al. 1990; Putnam 1987). The student submits a response to the question by typing and pressing the &amp;quot;Submit&amp;quot; button. The student's contribution is then segmented, parsed (Sekine and Grishman 1995) and sent through a rule-based utterance classifier. The classification process makes use of only the contribution text and part-of-speech tag provided by the parser.</Paragraph>
    <Paragraph position="6"> Mixed-initiative dialog starts with utterance classification and ends with dialog move generation, which can include question answering, repeating the question for the student, or just encouraging the student. Concurrently, the LSA module evaluates the quality of the student contributions, and in the tutor-initiative mode, the dialog move generator selects one or a combination of specific dialog moves that is both conversationally and pedagogically appropriate (Person et al 2000; Person et al. 2001). The Dialog Advancer Network (DAN) is the intermediary of dialog move generation in all instances, using information from the speech act classifier and LSA to select the next dialog move type and appropriate discourse markers. The dialog move generator selects the actual move. There are twelve types of dialog move: Pump, Hint, Splice, Prompt, Prompt Response, Elaboration, Summary, and five forms of immediate short-feedback (Graesser and Person 1994; Graesser et al. 1995; Person and Graesser 1999).</Paragraph>
  </Section>
class="xml-element"></Paper>