<?xml version="1.0" standalone="yes"?>
<Paper uid="C04-1141">
  <Title>Collocation Extraction Based on Modifiability Statistics</Title>
  <Section position="8" start_page="5" end_page="5" type="concl">
    <SectionTitle>
6 Conclusion
</SectionTitle>
    <Paragraph position="0"> We introduced a new, linguistically motivated measure of collocativity based on the property of limited modifiability and tested it on a large corpus with emphasis on German PP-verb combinations.</Paragraph>
    <Paragraph position="1"> We showed that our measure not only significantly outperforms the standard lexical association measures typically used for collocation extraction, but also yields a valuable by-product for the creation of collocation databases, viz. possible structural and lexical attributes of a collocation.</Paragraph>
    <Paragraph position="2"> Our measure defines the modifiability property in a linguistically simple way, by e.g. ignoring the internal make-up of lexical supplements associated with a collocation candidate. Hence, it may be worthwhile to investigate whether a more sophisticated approach, by e.g. taking into account internal POS types and their distribution etc., would improve our results even more. We may also consider other linguistic criteria (e.g., limited substitutability) to further refine our measure and to categorize already identified collocations.</Paragraph>
    <Paragraph position="3"> At the methodological level, our approach, although tested on German newspaper language data, is language-, structure-, and domain-independent.</Paragraph>
    <Paragraph position="4"> All it requires is some sort of shallow syntactic analysis, e.g., POS tagging and phrase chunking.</Paragraph>
    <Paragraph position="5"> Thus, in the future we plan to include other syntactic types of collocations, such as verb-object or verbobject-PP combinations, and also apply our methodology to other languages and domains, such as the biomedical field.</Paragraph>
    <Paragraph position="6"> Acknowledgements. We would like to thank our students, Sabine Demsar, Kristina Meller, and Konrad Feldmeier, for their excellent work as human collocation classifiers. This work was partly supported by DFG grant KL 640/5-1.</Paragraph>
  </Section>
class="xml-element"></Paper>