<?xml version="1.0" standalone="yes"?>
<Paper uid="W06-2504">
  <Title>What's in a name? The automatic recognition of metonymical location names.</Title>
  <Section position="6" start_page="30" end_page="31" type="concl">
    <SectionTitle>
4 Conclusions
</SectionTitle>
    <Paragraph position="0"> This paper has investigated two computational approaches to metonymy recognition that both in their own way are less complex than their competitors in the literature. The unsupervised algorithm in section 2 does not need any labelled training data; the supervised algorithm of Memory-Based Learning incorporates an extremely simple learning phase. Both approaches moreover have a clear relation to models of human behaviour.</Paragraph>
    <Paragraph position="1">  Sch&amp;quot;utze's (1998) approach is related to LSA, a model whose output correlates with human performance on a number of language tasks. Memory-Based Learning is akin to Case-Based Reasoning, which holds that people approach a problem by comparing it to similar instances in their memory.</Paragraph>
    <Paragraph position="2"> Rather than presenting a psycholinguistic critique of these approaches, this paper has investigated their ability to recognize metonymical loca-tion names. Not surprisingly, it was shown that the unsupervised approach is not yet a good basis for a robust metonymy recognition system. Nevertheless, it was often able to distinguish two clusters in the data that correlate with the literal and metonymical readings. It is striking that this is also the case for a set of mixed target words from the same category -- a type of data set that, to my knowledge, this algorithm had not yet been applied to. Memory-Based Learning, finally, proved to be a reliable way of recognizing metonymical words. Although this approach is much simpler than many competing algorithms, it produced state-of-the-art results, even without semantic information. null</Paragraph>
  </Section>
class="xml-element"></Paper>