<?xml version="1.0" standalone="yes"?>
<Paper uid="N06-2032">
  <Title>Story Segmentation of Brodcast News in English, Mandarin and Arabic</Title>
  <Section position="7" start_page="127" end_page="127" type="concl">
    <SectionTitle>
6 Conclusion
</SectionTitle>
    <Paragraph position="0"> In this paper we have presented results of our story boundary detection procedures on English, Mandarin, and Arabic Broadcast News from the TDT-4 corpus. All features are obtained automatically, except for the identity of the news show and the source language, information which is, however, available from the data itself, and could be automatically obtained. Our performance on TDT-4 BN appears to be better than previous work on earlier corpora of BN for English, and slightly worse than previous efforts on Mandarin, again for a different corpus. We believe our Arabic results to be the rst reported evaluation for BN in that language. One important observation from our study is that acoustic/prosodic features that correlate with story boundaries in English and in Mandarin, do not correlate with Arabic boundaries. Our further research will adress the study of vocal cues to segmentation in Arabic BN.</Paragraph>
  </Section>
class="xml-element"></Paper>