<?xml version="1.0" standalone="yes"?>
<Paper uid="W06-1634">
  <Title>Automatic Construction of Predicate-argument Structure Patterns for Biomedical Information Extraction</Title>
  <Section position="9" start_page="291" end_page="291" type="concl">
    <SectionTitle>
6 Conclusion
</SectionTitle>
    <Paragraph position="0"> We proposed the use of PASs to construct patterns as extraction rules, utilizing their ability to abstract syntactical variants with the same relation. In addition, we divided the patterns for generalization, and used matching results for SVM learning. In experiments on extracting of protein-protein interactions, we obtained 71.8% precision  and48.4%recallonasmallcorpusand64.0%precision and 83.8% recall estimated on a large text, whichdemonstratedtheobviousadvantagesofour method.</Paragraph>
  </Section>
class="xml-element"></Paper>