<?xml version="1.0" standalone="yes"?>
<Paper uid="J92-4002">
  <Title>Ambiguous Noun Phrases in Logical Form</Title>
  <Section position="12" start_page="458" end_page="459" type="concl">
    <SectionTitle>
8. Conclusion
</SectionTitle>
    <Paragraph position="0"> We began with the idea that LF, a compact intermediate level of representation derived only from syntactic and sentence-level information, is a valuable component in any computer model generating meaning for sentences. It can be used to postpone the determination of an unambiguous meaning for a sentence until the information required to select the intended meaning becomes available. Even though the meaning of the logical form for a sentence is ambiguous, it can be used by a contextual  Computational Linguistics Volume 18, Number 4 processing module to guide the refinement of its own meaning (as well as the meaning of other sentences' LFs).</Paragraph>
    <Paragraph position="1"> We have developed logical form representations for pronouns, definites, and indefinites and have demonstrated through a variety of detailed examples that these representations accurately model the linguistic behavior of the language (for a discussion of a wider variety of examples see Harper \[1989, 1990\]). We have proposed three computational constraints for using LF in a computational framework and have demonstrated that our LF conforms to these constraints. And finally, we used our representation in an implementation that was capable of processing nontrivial examples. There are two topics that we have only touched on in this paper: the extraction of information from context and the handling of syntactic ambiguity. Though LF constrains the information sought by a contextual module, it does not completely specify a strategy for locating and processing contextual information. A single parse tree can be automatically mapped into LE However, in syntactically ambiguous sentences (e.g., Fred saw the bird with his binoculars) more than one parse is possible. A solution is to allow multiple LFs for a sentence (though this would violate the compactness constraint). Another solution, which we are currently pursuing, is to associate LFs with a parse forest for the sentence (resulting in a compact representation for the parse trees and logical forms).</Paragraph>
  </Section>
class="xml-element"></Paper>