<?xml version="1.0" standalone="yes"?>
<Paper uid="E06-3009">
  <Title>Example-Based Metonymy Recognition for Proper Nouns</Title>
  <Section position="5" start_page="76" end_page="77" type="concl">
    <SectionTitle>
4 Conclusions and future work
</SectionTitle>
    <Paragraph position="0"> Inthis paper Ihave explored anexample-based approach to metonymy recognition. Memory-Based Learning does away with the complexity of current supervised metonymy recognition algorithms.</Paragraph>
    <Paragraph position="1"> Even without semantic information, it is able to give state-of-the-art results similar to those in the literature. Moreover, not only is the complexity of current learning algorithms unnecessary; the number of labelled training instances can be reduced drastically, too. I have argued that selective sam- null pling can help choose those instances that aremost helpful to the classifier. A few distance-based algorithms were able to drastically reduce the number of training instances that is needed for a given accuracy, both forthecountry and the organization names.</Paragraph>
    <Paragraph position="2"> If current metonymy recognition algorithms are to be used in a system that can recognize all possible metonymical patterns across a broad variety of semantic classes, it is crucial that the required number of labelled training examples be reduced.</Paragraph>
    <Paragraph position="3"> This paper has taken the first steps along this path and has set out some interesting questions for future research. This research should include the investigation of new features that can make classifiers more robust and allow us to measure their confidence more reliably. This confidence measurement can then also be used in semi-supervised learning algorithms, for instance, where the classifier itself labels the majority of training examples. Only with techniques such as selective sampling and semi-supervised learning can the knowledge acquisition bottleneck in metonymy recognition be addressed.</Paragraph>
  </Section>
class="xml-element"></Paper>