<?xml version="1.0" standalone="yes"?>
<Paper uid="N06-2047">
  <Title>Engineering Management The Chinese University of Hong Kong</Title>
  <Section position="3" start_page="185" end_page="185" type="intro">
    <SectionTitle>
2 Corpora
</SectionTitle>
    <Paragraph position="0"> We used the Remedia corpus (Hirschman et al., 1999) and ChungHwa corpus (Xu and Meng, 2005) in our experiments. The Remedia corpus contains 55 training stories and 60 testing stories (about 20K words). Each story contains 20 sentences on average and is accompanied by ve types of questions: who, what, when, where and why. The ChungHwa corpus contains 50 training stories and 50 test stories (about 18K words). Each story contains 9 sentences and is accompanied by four questions on average.</Paragraph>
    <Paragraph position="1"> Both the Remedia and ChungHwa corpora contain the annotation of NE, anaphor referents and answer sentences.</Paragraph>
  </Section>
class="xml-element"></Paper>