<?xml version="1.0" standalone="yes"?>
<Paper uid="W97-1206">
  <Title>Computing prosodic properties in a data-to-speech system</Title>
  <Section position="9" start_page="43" end_page="44" type="concl">
    <SectionTitle>
6 Conclusion
</SectionTitle>
    <Paragraph position="0"> To determine the prosodic properties of a sentence in a text, information about both sentence-internal and sentence-external syntax and semantics is needed. In Text-to-Speech this information has to be obtained through text analysis, whereas in Data-to-Speech reliable information of this kind is readily available.</Paragraph>
    <Paragraph position="1"> As a consequence, Data-to-Speech provides a better basis for using sophisticated prosody assignment rules than Text-to-Speech.</Paragraph>
    <Paragraph position="2"> The prosodic rules discussed in this paper are implemented in a generic Data-to-Speech system  called D2S. They make crucial use of both sentence-internal and sentence-external semantic and syntactic information, provided by the system in the form of a Discourse Model and parse trees of the generated sentences. The reliable and detailed character of this information makes it possible to assign prosodic markings which are reliable and detailed as well. This in turn allows for a close control of pros-odic realization, resulting in natural-sounding intonation. null</Paragraph>
  </Section>
class="xml-element"></Paper>