<?xml version="1.0" standalone="yes"?>
<Paper uid="J92-4002">
  <Title>Ambiguous Noun Phrases in Logical Form</Title>
  <Section position="4" start_page="421" end_page="422" type="intro">
    <SectionTitle>
3. Pronouns in Logical Form
</SectionTitle>
    <Paragraph position="0"> Pronouns are a source of underspecification in a sentence: the antecedent of a pronoun cannot be determined using syntactic information alone, but requires a combination of syntactic, semantic, and contextual information. We divide the process of determining the meaning of a pronoun into two phases. First, we provide the pronoun's LF, using only syntactic and sentence-level information. This LF constrains the possible antecedents to be those NPs that are consistent with this local information. The LF is also a flag indicating that the sentence is underspecified because more than one antecedent for the pronoun is possible. Later, when the antecedent is determined, a task that often requires contextual information found in surrounding sentences, we provide a way to update our LF to include this information.</Paragraph>
    <Paragraph position="1"> In the rest of this section, we discuss the linguistic behaviors of pronouns we want to model, introduce their LF representation, discuss how that representation is updated once contextual information isolates antecedents for the pronouns, and describe how the approach models verb phrase ellipsis (VPE).</Paragraph>
    <Section position="1" start_page="421" end_page="422" type="sub_section">
      <SectionTitle>
3.1 Pronouns: Linguistic Behavior
</SectionTitle>
      <Paragraph position="0"> Mary P. Harper Ambiguous Noun Phrases in Logical Form When a pronoun's intrasentential antecedent is a universal NP, that pronoun takes on the behavior of the universal's variable. Consider the sentence Fred showed every girl her picture. Given that the antecedent for her is every girl, the pronoun adopts the behavior of a variable bound by the universal quantifier of its antecedent, as shown in Example 2.</Paragraph>
      <Paragraph position="1"> Example 2 Fred/showed every girly herj picture.</Paragraph>
      <Paragraph position="2"> Vx (if (girl x) (show Fred (x's picture) x)) In contrast, if a pronoun has an intersentential antecedent, then that pronoun cannot act like a bound variable, since quantifiers do not have scope across sentences in English. Consider the following two sentences: Fred likes everyone. But, he doesn't return the sentiment. The pronoun in the second sentence cannot be bound by the quantifier corresponding to everyone in the first sentence. However, a quantified NP in one sentence can be the antecedent for a pronoun in another sentence, as in Fred likes everyone. But, they don't return the sentiment. In this case, they adopts the discourse entity for the group of individuals that everyone quantifies over. Webber (1978) discusses how to construct discourse entities for nonanaphoric NPs in a sentence (both for quantified NPs and definite NPs that are quantified over) once that sentence has been disambiguated. A discourse entity is a designator for the entity or set of entities the NP evokes in the discourse model of the speaker or hearer. If an NP in one sentence is the antecedent for a pronoun in another, the pronoun is replaced with the antecedent's discourse entity.</Paragraph>
      <Paragraph position="3"> Pronouns have also been classified as bound variable pronouns or referential pronouns. The antecedent for a bound variable pronoun occurs in the same sentence as the pronoun and the meaning of the pronoun is represented as a variable bound by the operator associated with its antecedent. In contrast, the meaning of a referential pronoun is the discourse entity evoked by its antecedent. The bound versus referential dichotomy divides the world of pronouns differently than does the intrasentential-intersentential dichotomy. Pronouns with intersentential antecedents are typically referential. 2 However, pronouns with intrasentential antecedents can be bound or referential (Webber 1978; Reinhart 1983).</Paragraph>
      <Paragraph position="4"> The bound-referential dichotomy doesn't cover the entire range of behaviors possible for pronouns. There is another category of pronoun that Evans (1980) dubs E-type pronouns and that appears to be a bound variable, but on closer inspection is not. Donkey sentences (originally noticed by Geach \[1962\]) can be used to demonstrate this difficulty, for example: Every miner who owns a donkey beats it. Given that every miner has scope over a donkey, the indefinite cannot be referential and the existential operator is blocked from binding the pronoun because of the scope island; quantified NPs embedded in a relative clause attached to an NP cannot bind pronouns outside of the relative clause environment. 3 However, a donkey can be the antecedent for the pronoun.</Paragraph>
    </Section>
  </Section>
class="xml-element"></Paper>