<?xml version="1.0" standalone="yes"?>
<Paper uid="W02-1028">
  <Title>A Bootstrapping Method for Learning Semantic Lexicons using Extraction Pattern Contexts</Title>
  <Section position="6" start_page="2" end_page="2" type="concl">
    <SectionTitle>
4 Conclusions
</SectionTitle>
    <Paragraph position="0"> Basilisk's bootstrapping algorithm exploits two ideas: (1) collective evidence from extraction patterns can be used to infer semantic category associations, and (2) learning multiple semantic categories simultaneously can help constrain the bootstrapping process. The accuracy achieved by Basilisk is substantially higher than that of previous techniques for semantic lexicon induction on the MUC-4 corpus, and empirical results show that both of Basilisk's ideas contribute to its performance. We also demon-Building: theatre store cathedral temple palace penitentiary academy houses school mansions Event: ambush assassination uprisings sabotage takeover incursion kidnappings clash shoot-out Human: boys snipers detainees commandoes extremists deserter narcoterrorists demonstrators cronies missionaries Location: suburb Soyapango capital Oslo regions cities neighborhoods Quito corregimiento Time: afternoon evening decade hour March weeks Saturday eve anniversary Wednesday Weapon: cannon grenade launchers rebomb car-bomb rifle pistol machineguns rearms  strated that learning multiple semantic categories simultaneously improves the meta-bootstrapping algorithm, which suggests that this is a general observation which may improve other bootstrapping algorithms as well.</Paragraph>
  </Section>
class="xml-element"></Paper>