<?xml version="1.0" standalone="yes"?>
<Paper uid="N04-1018">
  <Title>Detecting Structural Metadata with Decision Trees and Transformation-Based Learning</Title>
  <Section position="2" start_page="0" end_page="0" type="intro">
    <SectionTitle>
1 Introduction
</SectionTitle>
    <Paragraph position="0"> Automatic speech-to-text (STT) transcripts of spontaneous speech are often dif cult to comprehend even without the challenges arising from word recognition errors introduced by imperfect STT systems (Jones et al., 2003).</Paragraph>
    <Paragraph position="1"> Such transcripts lack punctuation that indicates clausal or sentential boundaries, and they contain a number of disuencies that would not normally occur in written language. Repeated words, hesitations such as um and uh , and corrections to a sentence in mid-stream are a normal part of conversational speech. These dis uencies are handled easily by human listeners (Shriberg, 1994), but their existence makes transcripts of spontaneous speech ill-suited for most natural language processing (NLP) systems developed for text, such as parsers or information extraction systems. Similarly, the lack of meaningful segmentation in automatically generated speech transcripts makes them problematic to use in NLP systems, most of which are designed to work at the sentence level. Detecting and removing dis uencies and locating sentential unit boundaries in spontaneous speech transcripts can improve their readability and make them more suitable for NLP. Automatically annotating discourse markers and other conversational llers is also likely to be useful, since proper handling is needed to follow the ow of conversation. Hence, the overall goal of our work is to detect such structural information in conversational speech using features generated by currently available speech processing systems and statistical machine learning tools.</Paragraph>
    <Paragraph position="2"> This paper is organized as follows. In Section 2, we describe the types of metadata that this work addresses, followed by a discussion of related prior work in Section 3. Section 4 describes the system architecture and details the algorithms and features used by our system.</Paragraph>
    <Paragraph position="3"> Section 5 discusses the experimental paradigm and results. Finally we provide a summary and directions for future work in Section 6.</Paragraph>
  </Section>
class="xml-element"></Paper>