<?xml version="1.0" standalone="yes"?>
<Paper uid="N06-2021">
  <Title>Initial Study on Automatic Identification of Speaker Role in Broadcast News Speech</Title>
  <Section position="3" start_page="0" end_page="0" type="intro">
    <SectionTitle>
TRECVID evaluations.1
</SectionTitle>
    <Paragraph position="0"> formation on video retrieval evaluations.</Paragraph>
    <Paragraph position="1"> In this paper, we develop algorithms for speaker role identification in broadcast news speech. Human transcription and manual speaker turn labels are used in this initial study. The task is then to classify each speaker's turn as anchor, reporter, or other. We use about 170 hours of speech for training and testing. Two approaches are evaluated, an HMM and a maximum entropy classifier. Our methods achieve about 80% accuracy for the three-way classification task, compared to around 50% when every speaker is labeled with the majority class label, i.e., anchor.2 The rest of the paper is organized as follows. Related work is introduced in Section 2. We describe our approaches in Section 3. Experimental setup and results are presented in Section 4. Summary and future work appear in Section 5.</Paragraph>
  </Section>
class="xml-element"></Paper>