<?xml version="1.0" standalone="yes"?>
<Paper uid="W06-1633">
  <Title>BESTCUT: A Graph Algorithm for Coreference Resolution</Title>
  <Section position="9" start_page="282" end_page="282" type="concl">
    <SectionTitle>
6 Conclusions
</SectionTitle>
    <Paragraph position="0"> We have proposed a novel coreference clusterization method that takes advantage of the efficiency and simplicity of graph algorithms. The approach is top-down and globally optimized, and takes into account cataphora resolution in addition to anaphora resolution. Our system compares favorably to two other implemented coreference systems and achieves state-of-the-art performance on the ACE Phase 2 corpus on true and detected mentions. We have also briefly described our mention detection system whose output we used in conjunction with the BESTCUT coreference system to achieve better results than when no mention information was available.</Paragraph>
  </Section>
class="xml-element"></Paper>