<?xml version="1.0" standalone="yes"?>
<Paper uid="N04-4035">
  <Title>Prosody-based Topic Segmentation for Mandarin Broadcast News</Title>
  <Section position="9" start_page="0" end_page="0" type="concl">
    <SectionTitle>
8 Conclusion and Future Work
</SectionTitle>
    <Paragraph position="0"> We have demonstrated the applicability of intonational prosodic features, specifically pitch, intensity, pause and duration, to the identification of topic boundaries in a tone language, Mandarin Chinese. We find highly significant decreases in pitch and intensity at topic final positions, and significant increases in word duration. Furthermore, these features in both local and contextualized form provide the basis for an effective decision tree classifier of boundary positions that does not use term similarity or cue phrase information, but only prosodic features. We also find that analogous to (Tur et al., 2001)'s work on an English story segmentation task, pause and pitch - both for the individual word and adjacency pair - play a crucial role; our findings for Chinese, however, identify a greater role played by intensity and durational contrasts.</Paragraph>
    <Paragraph position="1"> There is still substantial work to be done. We would like to integrate speaker identification for normalization and speaker change detection. We also plan to explore the integration of prosodic and textual features and investigate the identification of more fine-grained sub-topic structure, to provide more focused units for information retrieval, summarization, and anaphora resolution.</Paragraph>
  </Section>
class="xml-element"></Paper>