<?xml version="1.0" standalone="yes"?>
<Paper uid="W04-0816">
  <Title>Anselmo Pe~nas Dpto. Lenguajes y Sistemas Inform'aticos UNED, Spain anselmo@lsi.uned.es</Title>
  <Section position="7" start_page="0" end_page="0" type="concl">
    <SectionTitle>
6 Conclusion and work in progress
</SectionTitle>
    <Paragraph position="0"> We have shown the exemplar-based WSD system developed by UNED for the Senseval-3 lexical sample tasks. The general approach is based on the definition of a context space that becomes a flexible tool to prove quite different similarity measures between training contexts and new instances. We have shown that standard similarity measures improve their results applied inside this context space. We have established some criteria to instantiate this general approach and the resulting system has been evaluated at Senseval-3. The new similarity measure improves the disambiguation of nouns and obtains better overall results. The work in progress includes: * the study of new criteria to lead us to alternative measures, * the development of particular disambiguation strategies for verbs, nouns and adjectives, null * the inclusion of the dictionary information, and * the consideration of WordNet semantic relationships to extend the training corpus.</Paragraph>
  </Section>
class="xml-element"></Paper>