<?xml version="1.0" standalone="yes"?>
<Paper uid="H01-1055">
  <Title>RealizerSentencePlannerText Manager Dialog Natural Language Generation Planner Prosody Utterance User Utterance System Assigner TTS Natural Language Understanding ASR</Title>
  <Section position="7" start_page="2" end_page="2" type="concl">
    <SectionTitle>
6. CONCLUSION
</SectionTitle>
    <Paragraph position="0"> We have discussed how work in NLG can be applied in the development of dialog systems, and we have presented two approaches to using stochastic models and machine learning in NLG.</Paragraph>
    <Paragraph position="1"> Of course, the final justification for using a more sophisticated NLG architecture must come from user trials of an integrated system.</Paragraph>
    <Paragraph position="2"> However, we suspect that, as in the case of non-dialog NLG systems, the strongest arguments in favor of NLG often come from software engineering issues of maintainability and extensibility, which can be difficult to quantify in research systems.</Paragraph>
  </Section>
class="xml-element"></Paper>