<?xml version="1.0" standalone="yes"?>
<Paper uid="W06-0506">
  <Title>Sydney, July 2006. c(c)2006 Association for Computational Linguistics Taxonomy Learning using Term Specificity and Similarity</Title>
  <Section position="13" start_page="47" end_page="47" type="ackno">
    <SectionTitle>
6 Conclusion
</SectionTitle>
    <Paragraph position="0"> We have presented new taxonomy learning method with term similarity and specificity taken from domain-specific corpus. It can be applied to different domains as it is; and, if we have a syntactic parser available, to different languages. We analyzed the features used in previous researches in view of term specificity and similarity. In this analysis, we found that the features embed the characteristics of both conditions.</Paragraph>
    <Paragraph position="1"> Compared to previous approaches, our method has advantages in that we can use different features for term specificity and similarity. It makes easy to analyze errors in taxonomy learning step, whether the wrong relations are caused by specificity errors or by similarity errors. The main drawback of our method, as it is now, is that the effect of wrong located terms in upper level propagates to lower levels.</Paragraph>
    <Paragraph position="2"> Until now, researches on automatic ontology learning especially taxonomic relation showed very low precision. Human experts' intervention is inevitable in automatic learning process to make applicable taxonomy. Future work is to make new model where human experts and system work interactively in ontology learning process in order to balance cost and precision.</Paragraph>
  </Section>
class="xml-element"></Paper>