<?xml version="1.0" standalone="yes"?>
<Paper uid="W99-0619">
  <Title>Word Informativeness and Automatic Pitch Accent Modeling</Title>
  <Section position="4" start_page="0" end_page="148" type="intro">
    <SectionTitle>
POLICEMAN.&amp;quot;
</SectionTitle>
    <Paragraph position="0"> The capitalized words in the examples are accented. In (1), &amp;quot;man&amp;quot; is semantically empty relative to &amp;quot;kill&amp;quot;; therefore, the verb &amp;quot;kill&amp;quot; gets accented. However, in (2), &amp;quot;policeman&amp;quot; is semantically rich and is accented instead.</Paragraph>
    <Paragraph position="1"> However, different theories, not based on informativeness, were proposed to explain the above phenomenon. For example, Bresnan's (1971) explanation is based on syntactic function. She suggests that &amp;quot;man&amp;quot; in the above sentence does not get accented because &amp;quot;man&amp;quot; and other words like &amp;quot;guy&amp;quot; or &amp;quot;person&amp;quot; or &amp;quot;thing&amp;quot; form a category of  &amp;quot;semi-pronouns&amp;quot;. Counter-examples listed below raise more questions about the usefulness of semantic informativeness. The accent pattern in the following examples cannot be explaihed solely by semantic informativeness. null</Paragraph>
  </Section>
class="xml-element"></Paper>