<?xml version="1.0" standalone="yes"?>
<Paper uid="P05-3020">
  <Title>Automatic Part-of-Speech Induction from Text</Title>
  <Section position="7" start_page="79" end_page="79" type="concl">
    <SectionTitle>
5 Summary and Conclusions
</SectionTitle>
    <Paragraph position="0"> This work was inspired by previous work on word sense induction. The results indicate that part of speech induction is possible with good success based on the analysis of distributional patterns in text. The study also gives some insight how SVD is capable of significantly improving the results.</Paragraph>
    <Paragraph position="1"> Whereas in a previous paper (Rapp, 2004) we found that for word sense induction the local clustering of local vectors is more appropriate than the global clustering of global vectors, for part-of-speech induction our conclusion is that the situation is exactly the other way round, i.e. the global clustering of global vectors is more adequate (see footnote 1). This finding is of interest when trying to understand the nature of syntax versus semantics if expressed in statistical terms.</Paragraph>
  </Section>
class="xml-element"></Paper>