<?xml version="1.0" standalone="yes"?>
<Paper uid="W99-0101">
  <Title>O @ O O O O O @ O O O O O O O O @ O O O @ O O O O O O O @ O O O O O O O O O O An Integrated Approach to Reference and Presupposition Resolution</Title>
  <Section position="6" start_page="8" end_page="9" type="concl">
    <SectionTitle>
5 Conclusions
</SectionTitle>
    <Paragraph position="0"> We have described an integrated approach to resolving presuppositions, which includes pronominal and definite reference resolution Central to our approach is the maintenance of discourse structures, especially the QUD stack, which captures the hierarchical organization of the discourse. By identify8When all top-level ),-terms in a Wh-question have been replaced, it is interpreted as a polar question. ing the presuppositions associated with each kind of construction, and recovering presupposed information from a unified discourse information structure, the search space for relevant contextual information is restricted in a general way.</Paragraph>
    <Paragraph position="1"> Re~li~i-g the full potential of this approach to discourse structure requires recoL,,i~.in~ the question under discussion even when it is not stated explicitly as a question. We have found that requests and statements of need and desire should be coerced to questions, but a general implementation of this coercion process remains to be completed. An even more challenging problem involves inferring the domain goals of the user that are related to the question under discussion. As will be readily apparent to most researchers in NLP, these problems are intertwined with larger AI problems, such as plan recognition, which are beyond the scope of this paper, but must ultimately be solved or approximated in any realistic application. As noted in our discussion of domain restriction (SS4.3), it is sometimes possible to extract some explicitly mentioned content from the QUD which is relat~ to the user's domain goals, but we recoL,-i~e that this is only an apprarlmate solution.</Paragraph>
    <Paragraph position="2"> Although we have shown several examples where this approach to discourse structure is successful, a comprehensive empirical evaluation still needs to be performed to dete _rm_in_e how frequently particular presupposition problems occur in actual corpora, and to assess what proportion of the actual occurrences are resolved effectively by our algorithms.</Paragraph>
    <Paragraph position="3"> The first part of this evaluation could be addressed by searc~inf a corpus for expressions that trigger presuppositions, but assessing the actual performance of the algorithms on existing corpora would be more diMcult, because it would require the acquisition of background knowledge about other do- null mains than the hotel reservation application that we are currently working on.</Paragraph>
  </Section>
class="xml-element"></Paper>