<?xml version="1.0" standalone="yes"?>
<Paper uid="W04-0850">
  <Title>The Duluth Lexical Sample Systems in SENSEVAL-3</Title>
  <Section position="7" start_page="0" end_page="0" type="concl">
    <SectionTitle>
6 Conclusions
</SectionTitle>
    <Paragraph position="0"> This paper has described two lexical sample systems from the University of Minnesota, Duluth that participated in the SENSEVAL-3 exercise. We found that our supervised approach, Duluth-xLSS, fared reasonably well in a wide range of lexical sample tasks, thus suggesting that simple lexical features can serve as a firm foundation upon which to build a disambiguation system in a range of languages. The unsupervised approach of Duluth-ELSU to the English lexical sample task did not fare as well as the supervised approach, but performed at levels comparable to that attained by unsupervised systems in SENSEVAL-1 and SENSEVAL-2.</Paragraph>
  </Section>
class="xml-element"></Paper>