<?xml version="1.0" standalone="yes"?>
<Paper uid="P06-1099">
  <Title>You Can't Beat Frequency (Unless You Use Linguistic Knowledge) - A Qualitative Evaluation of Association Measures for Collocation and Term Extraction</Title>
  <Section position="7" start_page="790" end_page="790" type="concl">
    <SectionTitle>
5 Conclusions
</SectionTitle>
    <Paragraph position="0"> For lexical processing, the automatic identification of terms and collocations constitutes a research theme that has been dealt with by employing increasingly complex probabilistic criteria (ttest, mutual information, log-likelihood etc.). This trend is also reflected by their prominent status in standard textbooks on statistical NLP. The implicit justification in using these statistics-only metrics was that they would markedly outperform frequency of co-occurrence counting. We devised four qualitative criteria for explicitly testing this assumption. Using the best performing standard association measure (t-test) as a pars pro toto, our study indicates that the statistical sophistication does not pay off when compared with simple frequency of co-occurrence counting.</Paragraph>
    <Paragraph position="1"> This pattern changes, however, when probabilistic measures incorporate additional linguistic knowledge about the distributional properties of terms and the modifiability properties of collocations. Our results show that these augmented metrics reveal a marked difference compared to frequency of occurrence counts - to a larger degree with respect to automatic term recognition, to a slightly lesser degree for collocation extraction.</Paragraph>
  </Section>
class="xml-element"></Paper>