<?xml version="1.0" standalone="yes"?>
<Paper uid="A00-2034">
  <Title>Using Semantic Preferences to Identify Verbal Participation in Role Switching Alternations.</Title>
  <Section position="7" start_page="261" end_page="261" type="concl">
    <SectionTitle>
6 Conclusion
</SectionTitle>
    <Paragraph position="0"> We have discovered a significant relationship between the similarity of selectional preferences at the target slots, and participation in the causative and conative alternations. A threshold, such as the mean or median can be used to obtain a level of accuracy well above the baseline. A lemma based similarity score does not always indicate a significant relationship and generally produces a lower accuracy.</Paragraph>
    <Paragraph position="1"> There are patterns of diathesis behaviour among verb groups (Levin, 1993). Accuracy may be improved by considering severM alternations collectively, rather than in isolation. Complementary techniques to identify alternations, for example (Resnik, 1993), might be combined with ours.</Paragraph>
    <Paragraph position="2"> Although we have reported results on only two RSAS, our method is applicable to other such alternations. Furthermore, such application requires no human endeavour, apart from that required for evaluation. However, a considerably larger corpus would be required to overcome the sparse data problem for other RSA alternations.</Paragraph>
  </Section>
class="xml-element"></Paper>