<?xml version="1.0" standalone="yes"?>
<Paper uid="W02-0709">
  <Title>Interactive Chinese-to-English Speech Translation Based on Dialogue Management</Title>
  <Section position="3" start_page="2" end_page="2" type="concl">
    <SectionTitle>
4 Conclusion
</SectionTitle>
    <Paragraph position="0"> This paper describes a new paradigm for S2S translation system, which is based on DMM.</Paragraph>
    <Paragraph position="1"> According to the description we can see that the paradigm is of the following features: (1) The S2S translation is realized in the combination of direct translation engines and the interaction led by DMM.</Paragraph>
    <Paragraph position="2"> The interaction is not always brought into the role, and it only works when the former translation engines work failed.</Paragraph>
    <Paragraph position="3"> (2) The interaction is impersonative, targetoriented, and led by the system, not blind. The user does not need to correct all of the errors in the results of SR. He or she only needs to concern what the system asks.</Paragraph>
    <Paragraph position="4"> (3) The system can always give the results for an input speech despite of the ill-formed expressions and the worse recognition results.</Paragraph>
    <Paragraph position="5"> Although the whole experimental system is under construction, some preliminary results have been gained. Zong (2000c) reported the performance of the template-based translator; Xie (2002) reported the results of the robust parser for the Chinese spoken language; Xu (2001) presented the results of dialogue model; and so on. The results have made us confident to develop the practical S2S translation system based on the dialogue management. However, we are facing much hard work that involve the following aspects at least: null Develop the reasonable strategies and standards to evaluate the parsing results; null Design the effective templates to ask the user questions according the keywords and the concrete context; null Define the practical templates to generate the translation results; null Build the machine learning mechanism to enrich the knowledge base of the system.</Paragraph>
  </Section>
class="xml-element"></Paper>