<?xml version="1.0" standalone="yes"?>
<Paper uid="W02-1008">
  <Title>Combining Sample Selection and Error-Driven Pruning for Machine Learning of Coreference Rules</Title>
  <Section position="6" start_page="0" end_page="0" type="concl">
    <SectionTitle>
6 Conclusions
</SectionTitle>
    <Paragraph position="0"> We have examined three problems with recasting noun phrase coreference resolution as a classification task. To handle these problems, we presented a minimalist negative sample selection algorithm to reduce the skewness of the class distributions, and an automatic positive sample selection algorithm to select easy positive instances. In addition, our experiments indicate that the positive sample selection algorithm does not guarantee that hard instances can be entirely excluded. As a result, we proposed an error-driven rule pruning algorithm that can effectively enhance the precision of the system by dis14RULE-SELECT can be used in conjunction with any coreference scoring function. The MUC scorer is chosen here to facilitate comparison with previous results.</Paragraph>
    <Paragraph position="1"> carding rules that cause the ruleset to perform poorly with respect to the coreference scoring function.</Paragraph>
    <Paragraph position="2"> The resulting system outperformed the best MUC-6 and MUC-7 coreference systems as well as the best-performing learning-based system on the corresponding MUC data sets. Nevertheless, there is substantial room for improvement. For example, it is important to know how sensitive system performance is with respect to the size of the pruning corpus. In addition, although we use RIPPER as the underlying learning algorithm in our coreference system, we expect that the techniques described in this paper can be used in conjunction with other learning algorithms. We plan to explore this possibility in future work.</Paragraph>
  </Section>
class="xml-element"></Paper>