<?xml version="1.0" standalone="yes"?>
<Paper uid="W04-0850">
  <Title>The Duluth Lexical Sample Systems in SENSEVAL-3</Title>
  <Section position="2" start_page="0" end_page="0" type="intro">
    <SectionTitle>
1 Introduction
</SectionTitle>
    <Paragraph position="0"> The Duluth systems participated in various lexical sample tasks in SENSEVAL-3, using both supervised and unsupervised methodologies.</Paragraph>
    <Paragraph position="1"> The supervised lexical sample system that participated in SENSEVAL-3 is the Duluth3 (English) or Duluth8 (Spanish) system as used in SENSEVAL-2 (Pedersen, 2001b). It has been renamed for SENSEVAL-3 as Duluth-xLSS, where x is a one letter abbreviation of the language to which it is being applied, and LSS stands for Lexical Sample Supervised. The idea behind this system is to learn three bagged decision trees, one using unigram features, another using bigram features, and a third using co-occurrences with the target word as features. This system only uses surface lexical features, so it can be easily applied to a wide range of languages. For SENSEVAL-3 this system participated in the English, Spanish, Basque, Catalan, Romanian, and MultiLingual (English-Hindi) tasks.</Paragraph>
    <Paragraph position="2"> The unsupervised lexical sample system is based on the SenseRelate algorithm (Patwardhan et al., 2003) for word sense disambiguation. It is known as Duluth-ELSU, for English Lexical Sample Unsupervised. This system relies on measures of semantic relatedness in order to determine which sense of a word is most related to the possible senses of nearby content words. This system determines relatedness based on information extracted from the lexical database WordNet using the Word-Net::Similarity package. In SENSEVAL-3 this system was restricted to English text, although in future it and the WordNet::Similarity package could be ported to WordNets in other languages.</Paragraph>
    <Paragraph position="3"> This paper continues by describing our supervised learning technique which is based on the use of bagged decision trees, and then introduces the dictionary based unsupervised algorithm. We discuss our results from SENSEVAL-3, and conclude with some ideas for future work.</Paragraph>
  </Section>
class="xml-element"></Paper>