<?xml version="1.0" standalone="yes"?>
<Paper uid="W03-0502">
  <Title>Sub-event based multi-document summarization</Title>
  <Section position="3" start_page="3" end_page="3" type="intro">
    <SectionTitle>
2. Related Work
</SectionTitle>
    <Paragraph position="0"> Much work has preceded and informed this paper. Allan et al.'s (1998) work on summarizing novelty recognizes that news topics consist of a series of events - what we call &amp;quot;sub-events,&amp;quot; to distinguish the difference between a news topic and its sub-events. However, their method differs in its approach, which uses an algorithm to identify &amp;quot;novel&amp;quot; sentences, rather than the use of human judges. In other related work, sentences are either judged &amp;quot;on-topic&amp;quot; or &amp;quot;off-topic&amp;quot; (Allan et al., 2001a) (Allan et al., 2001b).</Paragraph>
    <Paragraph position="1"> Carbonell and Goldstein use Maximal Marginal Relevance (MMR) to identify &amp;quot;novel&amp;quot; information to improve query answering results, and they also apply this method to multiple-document summarization (Carbonell and Goldstein, 1997 and Goldstein, 1999). Success in the use of inter-judge agreement has led us to pursue the use of the current evaluation methods. However, this experiment differs from prior work in that we use judges to determine the relevance of sentences to sub-events rather than to evaluate summaries (Radev et al., 2000). Finally, McKeown et al. (1999), Hatzivassiloglou et al. (2001) and Boros et al. (2001) have shown the challenges and potential payoffs of using sentence clustering in extractive summarization.</Paragraph>
  </Section>
class="xml-element"></Paper>