<?xml version="1.0" standalone="yes"?>
<Paper uid="W03-1807">
  <Title>Extracting Multiword Expressions with A Semantic Tagger</Title>
  <Section position="5" start_page="3" end_page="3" type="concl">
    <SectionTitle>
6 Conclusion
</SectionTitle>
    <Paragraph position="0"> In this paper, we have shown that it is a practical way to extract MWEs using semantic field information. Since MWEs are lexical units carrying single semantic concepts, it is reasonable to consider the issue of MWE extraction as an issue of identifying word bundles depicting single semantic units. The main difficulty facing such an approach is that very few reliable automatic tools available for identifying lexical semantic units. However, a semantic field annotator, USAS, has been built in Lancaster University. Although it was not built aiming to the MWE extraction, we thought it might be very well suited for this purpose. Our experiment shows that the USAS tagger is indeed an efficient tool for MWE extraction.</Paragraph>
    <Paragraph position="1"> Nevertheless, the current semantic tagger does not provide a complete solution to the problem. During our experiment, we found that not all of the multiword units it collects are valid MWEs. An efficient algorithm is needed for distinguishing between free word combinations and relatively fixed, closely affiliated word bundles.</Paragraph>
  </Section>
class="xml-element"></Paper>