<?xml version="1.0" standalone="yes"?>
<Paper uid="N06-2045">
  <Title>Lycos Retriever: An Information Fusion Engine</Title>
  <Section position="11" start_page="179" end_page="179" type="ackno">
    <SectionTitle>
10 Related Work
</SectionTitle>
    <Paragraph position="0"> Although there have been previous systems that learned to identify and summarize web documents on a particular topic (Allen et al, 1996) without attempting to fuse them into a narrative structure, we are not aware of any project that attempts to generate coherent, narrative topical summaries by paragraph extraction and ordering. Much recent work focuses on multi-article summarization of news by sentence extraction and ordering (see for example, Columbia's well-known Newsblaster project and Michigan's NewsInEssence project).</Paragraph>
    <Paragraph position="1"> The latest DUC competition similarly emphasized sentence-level fusion of multi-document summaries from news text (DUC, 2005). One exception is the ArteQuaKt project (Kim et al, 2002), a prototype system for generating artist biographies from extracted passages and facts found on the Web aimed at different levels of readers (e.g. grade school versus university students). The Artequakt system was to use extracted text both as found and as generated from facts in a logical representation.</Paragraph>
    <Paragraph position="2"> It is not clear how far the ArteQuaKt project progressed. null Less legitimately, more and more &amp;quot;spam blogs&amp;quot; repackage snippets from search results or in other ways appropriate text from original sources into pages they populate with pay-per-click advertising. Retriever differs from such schemes in filtering out low value content and by making obscure sources visible.</Paragraph>
  </Section>
class="xml-element"></Paper>