<?xml version="1.0" standalone="yes"?>
<Paper uid="H05-1083">
  <Title>Proceedings of Human Language Technology Conference and Conference on Empirical Methods in Natural Language Processing (HLT/EMNLP), pages 660-667, Vancouver, October 2005. c(c)2005 Association for Computational Linguistics Multi-Lingual Coreference Resolution With Syntactic Features</Title>
  <Section position="7" start_page="665" end_page="666" type="concl">
    <SectionTitle>
6 Conclusions
</SectionTitle>
    <Paragraph position="0"> In this paper, linguistic knowledge is used to guide us to design features in maximum-entropy-based coreference resolution systems. In particular, we show how to compute a set of features to approximate the linguistic notions such as governing category and apposition, and how to compute the dependency features using syntactic parse trees. While the features are motivated by examining English data, we see signi cant improvements on both English and Arabic systems. Due to the language idiosyncrasy (e.g., pro-drops), we do not see the syntactic features change the Chinese system signi cantly.</Paragraph>
    <Paragraph position="1"> 3Ng and Cardie (2002) used a BINDING feature, but it is not clear from their paper how the feature was computed and what its impact was on their system.</Paragraph>
  </Section>
class="xml-element"></Paper>