<?xml version="1.0" standalone="yes"?>
<Paper uid="W02-0211">
  <Title>Discourse Processing for Explanatory Essays in Tutorial Applications</Title>
  <Section position="7" start_page="0" end_page="0" type="ackno">
    <SectionTitle>
Acknowledgments
</SectionTitle>
    <Paragraph position="0"> This research was supported by MURI grant N00014-00-1-0600 from ONR Cognitive Science and by NSF grant 9720359. We thank the entire NLT team for their many contributions in creating and building the Why-Atlas system. In particular we thank Michael Ringenberg, Maxim Makatchev, Uma Pappswamy and Michael B&amp;quot;ottner for their work with Tacitus-lite+ and the domain axioms and Roy Wilson for his work with the sentence-level realization module.</Paragraph>
    <Paragraph position="1"> 6An assumption cost of 1 means everything is assumed and a cost of 0 means that nothing was assumed.</Paragraph>
    <Paragraph position="2"> 7In these timing studies, we also did not allow the tutor to see the student input until the student had finished composing it. This was because our previous experiences with computer-mediated human tutoring have shown that some human tutors have a propensity for referring to something the student had started to write and then deleted. Our goal was to try to collect interactions that would be closer to those we expected with an intelligent tutoring system and was not primarily for comparing efficiency of a computer tutor to a human one.</Paragraph>
  </Section>
class="xml-element"></Paper>