<?xml version="1.0" standalone="yes"?>
<Paper uid="P06-1021">
  <Title>PCFGs with Syntactic and Prosodic Indicators of Speech Repairs</Title>
  <Section position="3" start_page="0" end_page="161" type="intro">
    <SectionTitle>
1 Introduction
</SectionTitle>
    <Paragraph position="0"> Speech repairs, as in example (1), are one kind of disfluent element that complicates any sort of syntax-sensitive processing of conversational speech.</Paragraph>
    <Paragraph position="1">  (1) and [ the first kind of invasion of ] the first  type of privacy seemed invaded to me The problem is that the bracketed reparandum region (following the terminology of Shriberg (1994)) is approximately repeated as the speaker The authors are very grateful for Eugene Charniak's help adapting his parser. We also thank the Center for Language and Speech processing at Johns Hopkins for hosting the summer workshop where much of this work was done. This material is based upon work supported by the National Science Foundation (NSF) under Grant No. 0121285. Any opinions, findings and conclusions or recommendations expressed in this material are those of the authors and do not necessarily reflect the views of the NSF.</Paragraph>
    <Paragraph position="2"> &amp;quot;repairs&amp;quot; what he or she has already uttered. This extra material renders the entire utterance ungrammatical--the string would not be generated by a correct grammar of fluent English. In particular, attractive tools for natural language understanding systems, such as Treebank grammars for written corpora, naturally lack appropriate rules for analyzing these constructions. One possible response to this mismatch between grammatical resources and the brute facts of disfluent speech is to make one look more like the other, for the purpose of parsing. In this separate-processing approach, reparanda are located through a variety of acoustic, lexical or string-based techniques, then excised before submission to a parser (Stolcke and Shriberg, 1996; Heeman and Allen, 1999; Spilker et al., 2000; Johnson and Charniak, 2004). The resulting parse tree then has the reparandum re-attached in a standardized way (Charniak and Johnson, 2001).</Paragraph>
    <Paragraph position="3"> An alternative strategy, adopted in this paper, is to use the same grammar to model fluent speech, disfluent speech, and their interleaving.</Paragraph>
    <Paragraph position="4"> Such an integrated approach can use syntactic properties of the reparandum itself. For instance, in example (1) the reparandum is an unfinished noun phrase, the repair a finished noun phrase. This sort of phrasal correspondence, while not absolute, is strong in conversational speech, and cannot be exploited on the separate-processing approach. Section 3 applies metarules (Weischedel and Sondheimer, 1983; McKelvie, 1998a; Core and Schubert, 1999) in recognizing these correspondences using standard context-free grammars.</Paragraph>
    <Paragraph position="5"> At the same time as it defies parsing, conversational speech offers the possibility of leveraging prosodic cues to speech repairs. Sec- null listener to identify the repair.</Paragraph>
    <Paragraph position="6"> tion 2 describes a classifier that learns to label prosodic breaks suggesting upcoming disfluency.</Paragraph>
    <Paragraph position="7"> These marks can be propagated up into parse trees and used in a probabilistic context-free grammar (PCFG) whose states are systematically split to encode the additional information.</Paragraph>
    <Paragraph position="8"> Section 4 reports results on Switchboard (Godfrey et al., 1992) and Fisher EARS RT04F data, suggesting these two features can bring about independent improvements in speech repair detection. Section 5 suggests underlying linguistic and statistical reasons for these improvements. Section 6 compares the proposed grammatical method to other related work, including state of the art separate-processing approaches. Section 7 concludes by indicating a way that string- and tree-based approaches to reparandum identification could be combined.</Paragraph>
  </Section>
class="xml-element"></Paper>