<?xml version="1.0" standalone="yes"?>
<Paper uid="W01-0716">
  <Title>Learning to identify animate references</Title>
  <Section position="8" start_page="6" end_page="6" type="concl">
    <SectionTitle>
6 Conclusions and future work
</SectionTitle>
    <Paragraph position="0"> In this paper, a two step method for animacy recognition was proposed. In the first step, it tries to determine the animacy of senses from WordNet on the basis of an annotated corpus. In the second step, this information is used by an instance based learning algorithm to determine the animacy of a noun. This area has been relatively neglected by researchers, therefore a comparison with other methods is difficult to make. The accuracy obtained is around 97%, more than 30% higher than that obtained by our previous system.</Paragraph>
    <Paragraph position="1"> Investigation of the results showed that in order to obtain accuracy close to 100%, several resources have to be used. As we point out in Section 4.2, a method which is able to weight the senses of a noun according to the text, and a named entity recogniser are necessary.</Paragraph>
    <Paragraph position="2"> The requirement for such components helps to emphasise the problematic nature of NP animacy recognition. We believe that such an investment should be made in order to go forward with this useful enterprise.</Paragraph>
  </Section>
class="xml-element"></Paper>