<?xml version="1.0" standalone="yes"?>
<Paper uid="W01-0513">
  <Title>Is Knowledge-Free Induction of Multiword Unit Dictionary Headwords a Solved Problem?</Title>
  <Section position="10" start_page="2" end_page="2" type="concl">
    <SectionTitle>
6 Conclusions
</SectionTitle>
    <Paragraph position="0"> This paper identifies several new results in the area of MWU-finding. We saw that MWU headword evaluations using WordNet provide similar results to those obtained from far more extensive web-based resources. Thus, one could safely use WordNet as a gold standard for future evaluations.</Paragraph>
    <Paragraph position="1"> We also noted that information-like algorithms, particularly Z-scores, SCP, and G242, seem to perform best at finding MRD headwords regardless of filtering mechanism, but that improvements are still needed. We proposed two new LSA-based approaches which attempted to address issues of non-compositionality and non-substitutivity.</Paragraph>
    <Paragraph position="2"> Apparently, either current algorithms already capture much non-compositionality or LSA-based models of non-compositionality are of little help.</Paragraph>
    <Paragraph position="3"> LSA does help somewhat as a model of substitutivity. However, LSA-based gains are small compared to the effort required to obtain them.</Paragraph>
  </Section>
class="xml-element"></Paper>