File Information

File: 05-lr/acl_arc_1_sum/cleansed_text/xml_by_section/relat/97/w97-1204_relat.xml

Size: 2,288 bytes

Last Modified: 2025-10-06 14:16:05

<?xml version="1.0" standalone="yes"?>
<Paper uid="W97-1204">
  <Title>Integrating Language Generation with Speech Synthesis Concept to Speech System</Title>
  <Section position="3" start_page="0" end_page="23" type="relat">
    <SectionTitle>
2 Related Work
</SectionTitle>
    <Paragraph position="0"> Recently, people have become more interested in developing CTS algorithms to improve the quality of synthesized speech. In (Prevost, 1995) and (Steedman, 1996), theme, rheme and contrast are used as important knowledge sources in determining accentual patterns. In (Davis and Hirschberg, 1988), given/new and topic structure are used to control intonational variation. Other CTS related research includes (Young and Fallside, 1979) and (Danlos et al., 1986). Most of the CTS systems developed to date have a closely integrated architecture. Because of this, CTS algorithms which map information from NLG to TTS parameters are system dependent.</Paragraph>
    <Paragraph position="1"> There is some related research in developing markup languages for TTS and speech transcription.</Paragraph>
    <Paragraph position="2"> The Speech Synthesis Markup Language( SSML) (Isard, 1995) is used as an interface for TTS. The motivation behind SSML is to overcome the difficulty that different TTS systems require different input format. No additional information is provided as input to TTS, but SSML provides a straightforward representation of existing prosodic features. This  representation is too simple for the purpose of integrating NLG and SS for CTS. There is almost no discourse, semantic or syntactic information in their representation, yet these are features one would expect as output from NLG and which should influence the prosody of speech.</Paragraph>
    <Paragraph position="3"> The Text Encoding Initiative (TEl) (Sperberg-McQueen and Burnard, 1993) provides a general guideline for transcribing spoken language using Standard Generalized Markup Language (SGML).</Paragraph>
    <Paragraph position="4"> SGML is an international standard for encoding electronic document for data interchange. Integrating two components in CTS is a specific SGML application. Therefore, it can't be addressed directly in SGML. But the design of SIML can be guided by TEI standards.</Paragraph>
  </Section>
class="xml-element"></Paper>
Download Original XML