<?xml version="1.0" standalone="yes"?>
<Paper uid="J81-2002">
  <Title>Relaxation Techniques for Parsing Grammatically Ill-Formed Input in Natural Language Understanding Systems 1</Title>
  <Section position="2" start_page="0" end_page="0" type="abstr">
    <SectionTitle>
1. Introduction
</SectionTitle>
    <Paragraph position="0"> Among the components included in a Natural Language Understanding (NLU) system is a grammar which specifies much of the linguistic structure of the utterances that can be expected. However, it is certain that inputs that are ill-formed with respect to the grammar will be received, both because people regularly form ungrammatical utterances and because there are a variety of forms that cannot be readily included in current grammatical models and are hence &amp;quot;extragrammatical&amp;quot;. These might be rejected, but as Wilks (1976) stresses, &amp;quot;... understanding requires, at the very least, ... some attempt to interpret, rather than merely reject, what seem to be ill-formed utterances.&amp;quot; This paper investigates several language phenomena commonly considered ungrammatical or extra-grammatical and discusses techniques directed at integrating them as much as possible into the conventional framework of grammatical processing performed by NLU systems. A &amp;quot;normative&amp;quot; grammar is assumed l This paper is a revised and extended version of a paper &amp;quot;Ungrammaticality and Extra-Grammaticality in Natural Language Understanding Systems&amp;quot; presented at the 17th Annual Meeting of the Association for Computational Linguistics, San Diego, California, August, 1979.</Paragraph>
    <Paragraph position="1"> which specifies the structure of well-formed inputs.</Paragraph>
    <Paragraph position="2"> Rules that are both manually added to the original grammar and, by relaxing the constraints of the grammar, automatically constructed during parsing analyze the ill-formed input. The ill-formedness is shown at the completion of a parse by an indication of its deviance from a fully grammatical structure. We have been able to accomplish this while preserving the structural characteristics of the original grammar and its inherent efficiency.</Paragraph>
    <Paragraph position="3"> The Augmented Transition Network (ATN) model was chosen as the tool in which to express our ideas, and a basic understanding of this model is assumed throughout the paper. 2 However, we believe our ideas are more general in scope than the ATN implementation may suggest. Similar ideas have recently been integrated into the Augmented Phrase Structure Grammar model by Miller, Heidorn, and Jensen (1981).</Paragraph>
    <Paragraph position="4"> Some of the phenomena discussed have been considered previously in particular NLU systems, as, for example, the ellipsis handling in LIFER described in Hendrix, Sacerdoti, Sagalowicz, and Slocum (1978).</Paragraph>
    <Paragraph position="5"> Similar methods for processing conjunction have been 2 For a thorough introduction to the ATN formalism, see Bates (1978).</Paragraph>
    <Paragraph position="6"> Copyright 1981 by the Association for Computational Linguistics. Permission to copy without fee all or part of this material is granted provided that the copies are not made for direct commercial advantage and the Journal reference and this copyright notice are included on the first page. To copy otherwise, or to republish, requires a fee and/or specific permission. 0362-613X/81/020099-10501.00 American Journal of Computational Linguistics, Volume 7, Number 2, April-June 1981 99 Stan C. Kwasny and Norman K. Sondheimer Relaxation Techniques for Parsing Ill-Formed Input used in the LUNAR system as discussed in Woods (1973). In linguistic studies, Chomsky (1964) and Katz (1964), among others, have considered the treatment of ungrammaticality within the Transformational Generative model. Contemporary studies closest to ours are those of Weischedel and Black (1980) and Hayes and Mouradian (1980). Similarities will be pointed out as our techniques are presented. The present study is distinguished by the range of phenomena considered, structural and efficiency goals, and the inclusion of techniques in one uniform framework.</Paragraph>
    <Paragraph position="7"> This paper surveys these problems, discusses mechanisms aimed at solving them, and discusses how these mechanisms are used. At the end, some limitations are discussed and extensions suggested. Unless otherwise noted, all ideas have been tested through implementation. A more detailed discussion of all points may be found in Kwasny (1980).</Paragraph>
  </Section>
class="xml-element"></Paper>