<?xml version="1.0" standalone="yes"?>
<Paper uid="A00-1040">
  <Title>Using Corpus-derived Name Lists for Named Entity Recognition</Title>
  <Section position="7" start_page="294" end_page="294" type="concl">
    <SectionTitle>
6 Conclusion
</SectionTitle>
    <Paragraph position="0"> This paper explored the role of lists of names in NE recognition, comparing hand-crafted and corpus-derived lists. It was shown that, under certain conditions, corpus-derived lists outperform hand-crafted ones. Also, supplementing hand-crafted lists with corpus-based ones often improves their performance.</Paragraph>
    <Paragraph position="1"> The reported method was more effective for the ORGANIZATION and LOCATION classes of names than for PERSON, which was attributed to the fact that reportage of these names does not change as much over time in broadcast news.</Paragraph>
    <Paragraph position="2"> The method reported here achieves 87% Fmeasure, 5% less than the best performing system in the HUB4 evaluation. However, it should be remembered that this technique uses only a simple application of internal evidence.</Paragraph>
  </Section>
class="xml-element"></Paper>