<?xml version="1.0" standalone="yes"?>
<Paper uid="A88-1008">
  <Title>HANDLING SCOPE AMBIGUITIES IN ENGLISH</Title>
  <Section position="12" start_page="63" end_page="63" type="concl">
    <SectionTitle>
CONCLUSION
</SectionTitle>
    <Paragraph position="0"> This paper has described some features of a program designed to handle scope ambiguities in English. Some of the more problematic issues which were encountered during the designing of the program were selected for discussion: the choice of logical representation, the seeping of coordinated expressions, the choice of a strategy for selecting preferred scope orderings and the determination of a set of domain-independent heuristics. The program is currently being extended to include a wider range of lexical types and input expressions and the heuristics are being improved. Following this, it is hoped to incorporate some simple types of domainand discourse-dependent knowledge into the program, in particular knowledge about expected relations among objects in a given domain and a simple discourse focus structure.</Paragraph>
    <Paragraph position="1"> The selection of preferred scope orderings depends on the complex interaction of linguistic and context-dependent knowldege. It would be a considerable advantage to be able to factor out the contributions of different types of knowledge required and then at some later time to combine them. One conclusion of this work is that there is a body of largely domain-independent knowledge which can play an important, and at times decisive, role in the disambiguation of scope.</Paragraph>
    <Paragraph position="2"> Such knowledge is most useful when it indicates a very strong or absolute preference for one reading.</Paragraph>
    <Paragraph position="3"> Absolute preferences typically occur with operators such as any or both and with distributive quantifiers or and inside strong clausal trap or inside a coordinator. Very strong preferences may occur with operators such as few, no or each, with preposed or topicalized operators and with operators inside prepositional phrases. When the domain-independent heuristics do not provide a strong preference for one reading, they may still serve as a useful guide guide for the later application of pragmatic knowledge. This is commonly the case when indefinites are present, as the &amp;quot;specificity&amp;quot; of indefinites is mainly context-dependent.</Paragraph>
    <Paragraph position="4"> A number of problems have not been discussed here because they remain unresolved. These include: the scoping of quantifiers relative to tense and opaque operators, the logical representation and scoping of generics, the treatment of pronouns not embedded within their quantifier antecedents, non-local problems such as the preference for &amp;quot;symmetric&amp;quot; readings, the use of stray words, such as together and both (as an adverb), which provide important clues for preferred scope relations and the difficult problem of combining linguistic and context-dependent heuristic knowledge.</Paragraph>
  </Section>
class="xml-element"></Paper>