<?xml version="1.0" standalone="yes"?>
<Paper uid="W02-0812">
  <Title>Evaluating the Effectiveness of Ensembles of Decision Trees in Disambiguating Senseval Lexical Samples</Title>
  <Section position="9" start_page="0" end_page="0" type="concl">
    <SectionTitle>
8 Conclusion
</SectionTitle>
    <Paragraph position="0"> This paper analyzes the performance of the Duluth3 and Duluth8 systems that participated in the English and Spanish lexical sample tasks in SENSEVAL2. We find that an ensemble offers very limited improvement over individual decision trees based on lexical features. Co-occurrence decision trees are more accurate than bigram or unigram decision trees, and are nearly as accurate as the full ensemble.</Paragraph>
    <Paragraph position="1"> This is an encouraging result, since the number of co-occurrence features is relatively small and easy to learn from compared to the number of bigram or unigram features.</Paragraph>
  </Section>
class="xml-element"></Paper>