<?xml version="1.0" standalone="yes"?>
<Paper uid="W04-0827">
  <Title>GAMBL, Genetic Algorithm Optimization of Memory-Based WSD</Title>
  <Section position="10" start_page="0" end_page="0" type="concl">
    <SectionTitle>
5 Conclusion
</SectionTitle>
    <Paragraph position="0"> From previous research on memory-based WSD, we learned that both feature selection, algorithm parameter settings, and their interaction, play an important role in accuracy, and that good selections and settings do not generalize over different word experts. These should therefore be optimized individually. We showed in this paper that using Genetic Algorithms and TIMBL, this complex multiple optimization problem can nevertheless be achieved, even for the AW task in which 3433 word experts have to be optimized.</Paragraph>
    <Paragraph position="1"> Compared with our previous system (Hoste et al., 2002), using chunks and grammatical relations as a source of information is an innovation. This in-</Paragraph>
  </Section>
class="xml-element"></Paper>