<?xml version="1.0" standalone="yes"?>
<Paper uid="W03-1015">
  <Title>Bootstrapping Coreference Classifiers with Multiple Machine Learning Algorithms</Title>
  <Section position="9" start_page="6" end_page="6" type="concl">
    <SectionTitle>
7 Conclusions
</SectionTitle>
    <Paragraph position="0"> We have proposed a single-view, multi-learner bootstrapping algorithm for coreference resolution and shown empirically that the algorithm is a better alternative to the Blum and Mitchell co-training algorithm for this task for which no natural feature split has been found. In addition, we have investigated an example ranking method for bootstrapping that, unlike Blum and Mitchell's rank-by-confidence method, can potentially alleviate the problem of performance deterioration due to the pollution of the labeled data in the course of bootstrapping.</Paragraph>
  </Section>
class="xml-element"></Paper>