<?xml version="1.0" standalone="yes"?>
<Paper uid="J83-3001">
  <Title>Recovery Strategies for Parsing Extragrammatical Language 1</Title>
  <Section position="9" start_page="0" end_page="0" type="concl">
    <SectionTitle>
6. Concluding Remarks
</SectionTitle>
    <Paragraph position="0"> Any practical natural language interface must be capable of dealing with a wide range of extragrammatical 144 American Journal of Computational Linguistics, Volume 9, Numbers 3-4, July-December 1983 Jaime G. Carbonell and Philip J. Hayes Recovery Strategies for Parsing Extrammatical Language input. This paper has proposed a taxonomy of the prevalent forms of extragrammaticality in real language use and presented recovery strategies for many of them. We also discussed how well various approaches to parsing could support the recovery strategies, and concluded that case frame instantiation provided the best framework among the commonly used parsing methodologies.</Paragraph>
    <Paragraph position="1"> At a more general level, we argued that the superiority of case frame instantiation over other parsing methodologies for robust parsing is due to how well it satisfies four parsing characteristics that are important for many of the recovery strategies that we described:  top-down as well as bottom-up.</Paragraph>
    <Paragraph position="2"> We claimed that while case frame instantiation satisfies these desiderata better than any other commonly used parsing methodology, it was possible to do even better by using a multi-strategy approach in which case frame instantiation was just one member (albeit a very important one) of a whole array of parsing and recovery strategies. We described some experiments that led us to this view and outlined a parsing methodology, entity-oriented parsing, that we believe will support a multi-strategy approach.</Paragraph>
    <Paragraph position="3"> It is our hope that by pursuing lines of research leading to parsers that maximize the characteristics listed above, we can approach, in semantically limited domains, the extraordinary degree of robustness in language recognition exhibited by human beings, and gain some insights into how robustness might be achieved in more general language settings.</Paragraph>
  </Section>
class="xml-element"></Paper>