<?xml version="1.0" standalone="yes"?>
<Paper uid="W06-3503">
  <Title>Understanding Complex Natural Language Explanations in Tutorial Applications[?]</Title>
  <Section position="6" start_page="22" end_page="23" type="concl">
    <SectionTitle>
5 Conclusion
</SectionTitle>
    <Paragraph position="0"> In this paper, we discussed an application that integrates a hybrid of semantic parsers and a symbolic reasoner with a statistical classifier to analyze student explanations. We attempted to address the problem that the leap made by statistical classifiers from NL to a feasible classification is too big since too many details of what was actually said by the student are lost. On the other hand, we showed that the hybrid semantic parsers allow for a slightly smaller leap by mapping to a symbolic representation that is sufficient for domain reasoning. Using deductive closure of problem givens and buggy assumptions, the correctness and completeness analyzer allows us to reason about the correctness of student statements that cannot be confidently classified statistically. Although formal and informal language expressions have unique underlying semantics, we attempt to paraphrase informal NL into formal NL by using the forward-chaining rules involved in creating the deductive closure for a problem from its givens. Our current symbolic representation is still too coarse to distinguish some fine nuances allowed by the domain of mechanics. We conjecture that extending our knowledge representation with more language-specific predicates would allow us to represent more fine-grained differences in student statements while still allowing feasible reasoning with the ATMS.</Paragraph>
  </Section>
class="xml-element"></Paper>