<?xml version="1.0" standalone="yes"?>
<Paper uid="W97-1206">
  <Title>Computing prosodic properties in a data-to-speech system</Title>
  <Section position="4" start_page="0" end_page="0" type="intro">
    <SectionTitle>
1 Introduction
</SectionTitle>
    <Paragraph position="0"> The central topic of this paper is the problem of computing the prosodic properties of sentences generated ill Data-to-Speech systems (i.e., systems which present data in the form of a spoken monologue sometimes also called 'Concept-to-Speech' systems).</Paragraph>
    <Paragraph position="1"> We propose a set of rules for the assignment of prosodic properties which take an explicit discourse model into account. In contrast to Text-to-Speech systems (and more generally, systems which require linguistic analysis of the input), explicit discourse models can be reliably constructed in Data-to-Speech systems (and more generally, in systems which generate natural language from data), so that a more natural prosody can be achieved.</Paragraph>
    <Paragraph position="2"> The rules for prosody assignment described in this paper are used in the language generation compon*Authors Theune and Klabbers carried out this research within the framework of the Priority Programme Language and Speech Technology (TST). The TSTprogramme is sponsored by NWO (Dutch Organization for Scientific Research).</Paragraph>
    <Paragraph position="3"> eat ofD2S, a generic system for the creation of Data-to-Speech systems. The method for natural language generation implemented in D2S is hybrid in nature (Reiter, 1995); (Coch, 1996). It is a particular mixture of (syntactic) template-based techniques and full natural language generation, described in more detail in Klabbers et al. (1997a). A variety of Data-to-Speech systems have been and are being developed on the basis of D2S. Examples are the Dial Your Disc (DYD)-system, which presents information in English about Mozart compositions (Deemter et al., 1994); (Collier and Landsbergen, 1995), and the GoalGetter system, which presents spoken monologues in Dutch about the course and the result of a football game (Klabbers et al., 1997b). In this paper, we illustrate the prosodic rules used in D2S with examples from GoalGetter.</Paragraph>
    <Paragraph position="4"> After a brief description and illustration of the general architecture of D2S, we describe in detail how the prosodic component of D2S computes the prosodic properties of the generated sentences. Then we discuss how the resulting prosodic annotations are used in the various speech output techniques employed in D2S. We end with some remarks about evaluation of the prosodic rules and a conclusion.</Paragraph>
  </Section>
class="xml-element"></Paper>