<?xml version="1.0" standalone="yes"?>
<Paper uid="W06-0503">
  <Title>Max-Planck-Institute for Computer Science</Title>
  <Section position="7" start_page="23" end_page="23" type="concl">
    <SectionTitle>
4 Conclusion and Outlook
</SectionTitle>
    <Paragraph position="0"> We addressed the problem of automatically extracting instances of arbitrary binary relations from natural language text. The key novelty of our approach is to apply a deep syntactic analysis to this problem. We have implemented our approach and showed that our system LEILA outperforms existing competitors.</Paragraph>
    <Paragraph position="1"> Our current implementation leaves room for future work. For example, the linkages allow for more sophisticated ways of resolving anaphoras or matching patterns. LEILA could learn numerous interesting relations (e.g. country / president or isAuthorOf) and build up an ontology from the results with high confidence.</Paragraph>
    <Paragraph position="2"> LEILA could acquire and exploit new corpora on its own (e.g., it could read newspapers) and it could use its knowledge to acquire and structure its new knowledge more efficiently. We plan to exploit these possibilities in our future work.</Paragraph>
    <Section position="1" start_page="23" end_page="23" type="sub_section">
      <SectionTitle>
4.1 Acknowledgements
</SectionTitle>
      <Paragraph position="0"> We would like to thank Eugene Agichtein for his caring support with Snowball. Furthermore, Johanna V&amp;quot;olker and Philipp Cimiano deserve our sincere thanks for their unreserved assistance with their system.</Paragraph>
    </Section>
  </Section>
class="xml-element"></Paper>