<?xml version="1.0" standalone="yes"?>
<Paper uid="W98-0609">
  <Title>Automatic Collection and Analysis of GermanCompounds</Title>
  <Section position="11" start_page="65" end_page="66" type="concl">
    <SectionTitle>
8. Conclusions
</SectionTitle>
    <Paragraph position="0"> Our goal in this effort has been to evaluate and, ultimately, to use for practical ends the analysis of large-scale German corpora in order to determine a morphological property of individual German noun stems -- the choice of Linker element used in compounding.</Paragraph>
    <Paragraph position="1"> Our results support the strategy of using large-scale natural language corpora as a source for automatic processing and as a means to gather specific lexical information. While linker information is sparsely distributed across the corpora we have studied, the largely automatic character of our search allows us to have increasingly certain information about this property.</Paragraph>
    <Paragraph position="2"> 16 The number of noun entries for which any distribution is calculated is, of course, dependent upon the corpus processed. Every step of processing described in this paper is fully automated, so that an arbitrarily large corpus can be processed, limited only by computational resources.</Paragraph>
  </Section>
class="xml-element"></Paper>