<?xml version="1.0" standalone="yes"?>
<Paper uid="H05-1115">
  <Title>Using Random Walks for Question-focused Sentence Retrieval</Title>
  <Section position="14" start_page="921" end_page="921" type="concl">
    <SectionTitle>
4 President Clinton's national security adviser, 0.0124 N
</SectionTitle>
    <Paragraph position="0"> Samuel Berger, has provided his Russian counterpart with a written summary of what U.S. naval and intelligence officials believe caused the nuclear-powered submarine Kursk to sink last month in the Barents Sea, officials said Wednesday.</Paragraph>
    <Paragraph position="1"> 5 There has been no final word on what caused 0.0123 N the submarine to sink while participating in a major naval exercise, but Defense Minister Igor Sergeyev said the theory that Kursk may have collided with another object is receiving increasingly concrete confirmation.</Paragraph>
    <Paragraph position="2"> Table 9: Top ranked sentences using the LR[0.20,0.95] system on the question &amp;quot;What caused the Kursk to sink?&amp;quot; answers from the retrieved sentences. In this case, the sentences selected by our system would be sent to an answer identification component for further processing. As discussed in Section 2, our goal was to develop a topic-sensitive version of LexRank and to use it to improve a baseline system, which had previously been used successfully for query-based sentence retrieval (Allan et al., 2003). In terms of this task, wehave shown that over a large set of unaltered questions written by our annotators, LexRank can, on average, outperform the baseline system, particularly in terms of TRDR scores.</Paragraph>
  </Section>
class="xml-element"></Paper>