<?xml version="1.0" standalone="yes"?>
<Paper uid="H01-1003">
  <Title>Advances in Meeting Recognition</Title>
  <Section position="6" start_page="0" end_page="0" type="concl">
    <SectionTitle>
5. CONCLUSIONS
</SectionTitle>
    <Paragraph position="0"> In this paper we have reviewed work on speech recognition systems applied to data from human-to-human interaction as encountered in meetings. The task is very challenging with error rates of 5-10 times higher than read speech (BN F0-condition) which basically results from degraded recording conditions, highly topic dependent dictionary and language models, as well as from the informal, conversational multi-party scenario. Our experiments using different training data, language modeling interpolation, adaptation and signal mapping yield more than 20% relative improvements in error rate.</Paragraph>
  </Section>
class="xml-element"></Paper>