<?xml version="1.0" standalone="yes"?>
<Paper uid="C04-1158">
  <Title>Efficient Confirmation Strategy for Large-scale Text Retrieval Systems with Spoken Dialogue Interface</Title>
  <Section position="6" start_page="5" end_page="5" type="concl">
    <SectionTitle>
5 Conclusion
</SectionTitle>
    <Paragraph position="0"> We described an appropriate confirmation strategy for large-scale text retrieval systems with a spoken dialogue interface. We introduced two measures, relevance score and significance score, for ASR results. The measures are useful to control confirmation efficiently for portions including either ASR errors or redundant expressions.</Paragraph>
    <Paragraph position="1"> The portions to be confirmed are determined  We used a word-level CM only because defining semantic categories for content words is required to calculate the concept-level CM. Because the semantic category corresponded to items in a relational database, we cannot use the concept-level CM in this task.</Paragraph>
    <Paragraph position="2">  using information that is automatically derived from the target knowledge base, such as a statistical language model, tf*idf values, and retrieval results. An experimental evaluation shows that our method can efficiently generate confirmations for better task achievement compared with that using a conventional confidence measure of the ASR. Our method is not dependent on the software support task, and expected to be applicable to general text retrieval tasks.</Paragraph>
  </Section>
class="xml-element"></Paper>