<?xml version="1.0" standalone="yes"?>
<Paper uid="W04-0807">
  <Title>The SENSEVAL-3 English Lexical Sample Task</Title>
  <Section position="3" start_page="0" end_page="0" type="intro">
    <SectionTitle>
3 Participating Systems
</SectionTitle>
    <Paragraph position="0"> 27 teams participated in this word sense disambiguation task. Tables 2 and 3 list the names of the participating systems, the corresponding institutions, and the name of the first author - which can be used as reference to a paper in this volume, with more detailed descriptions of the systems and additional analysis of the results.</Paragraph>
    <Paragraph position="1"> There were no restrictions placed on the number of submissions each team could make. A total number of 47 submissions were received for this task.</Paragraph>
    <Paragraph position="2"> Tables 2 and 3 show all the submissions for each team, gives a brief description of their approaches, and lists the precision and recall obtained by each system under fine and coarse grained evaluations.</Paragraph>
    <Paragraph position="3"> The precision/recall baseline obtained for this task under the &amp;quot;most frequent sense&amp;quot; heuristic is 55.2% (fine grained) and 64.5% (coarse grained). The performance of most systems (including several unsupervised systems, as listed in Table 3) is significantly higher than the baseline, with the best system performing at 72.9% (79.3%) for fine grained (coarse grained) scoring.</Paragraph>
    <Paragraph position="4"> Not surprisingly, several of the top performing systems are based on combinations of multiple classifiers, which shows once again that voting schemes that combine several learning algorithms outperform the accuracy of individual classifiers.</Paragraph>
  </Section>
class="xml-element"></Paper>