<?xml version="1.0" standalone="yes"?>
<Paper uid="N04-1038">
  <Title>Unsupervised Learning of Contextual Role Knowledge for Coreference Resolution</Title>
  <Section position="2" start_page="0" end_page="0" type="intro">
    <SectionTitle>
1 Introduction
</SectionTitle>
    <Paragraph position="0"> The problem of coreference resolution has received considerable attention, including theoretical discourse models (e.g., (Grosz et al., 1995; Grosz and Sidner, 1998)), syntactic algorithms (e.g., (Hobbs, 1978; Lappin and Leass, 1994)), and supervised machine learning systems (Aone and Bennett, 1995; McCarthy and Lehnert, 1995; Ng and Cardie, 2002; Soon et al., 2001). Most computational models for coreference resolution rely on properties of the anaphor and candidate antecedent, such as lexical matching, grammatical and syntactic features, semantic agreement, and positional information.</Paragraph>
    <Paragraph position="1"> The focus of our work is on the use of contextual role knowledge for coreference resolution. A contextual role represents the role that a noun phrase plays in an event or relationship. Our work is motivated by the observation that contextual roles can be critically important in determining the referent of a noun phrase. Consider the following sentences: (a) Jose Maria Martinez, Roberto Lisandy, and Dino Rossy, who were staying at a Tecun Uman hotel, were kidnapped by armed men who took them to an  unknown place.</Paragraph>
    <Paragraph position="2"> (b) After they were released...</Paragraph>
    <Paragraph position="3"> (c) After they blindfolded the men...</Paragraph>
    <Paragraph position="4">  In (b) &amp;quot;they&amp;quot; refers to the kidnapping victims, but in (c) &amp;quot;they&amp;quot; refers to the armed men. The role that each noun phrase plays in the kidnapping event is key to distinguishing these cases. The correct resolution in sentence (b) comes from knowledge that people who are kidnapped are often subsequently released. The correct resolution in sentence (c) depends on knowledge that kidnappers frequently blindfold their victims.</Paragraph>
    <Paragraph position="5"> We have developed a coreference resolver called BABAR that uses contextual role knowledge to make coreference decisions. BABAR employs information extraction techniques to represent and learn role relationships. Each pattern represents the role that a noun phrase plays in the surrounding context. BABAR uses unsupervised learning to acquire this knowledge from plain text without the need for annotated training data. Training examples are generated automatically by identifying noun phrases that can be easily resolved with their antecedents using lexical and syntactic heuristics. BABAR then computes statistics over the training examples measuring the frequency with which extraction patterns and noun phrases co-occur in coreference resolutions.</Paragraph>
    <Paragraph position="6"> In this paper, Section 2 begins by explaining how contextual role knowledge is represented and learned.</Paragraph>
    <Paragraph position="7"> Section 3 describes the complete coreference resolution model, which uses the contextual role knowledge as well as more traditional coreference features. Our coreference resolver also incorporates an existential noun phrase recognizer and a Dempster-Shafer probabilistic model to make resolution decisions. Section 4 presents experimental results on two corpora: the MUC-4 terrorism corpus, and Reuters texts about natural disasters. Our results show that BABAR achieves good performance in both domains, and that the contextual role knowledge improves performance, especially on pronouns. Finally, Section 5 explains how BABAR relates to previous work, and Section 6 summarizes our conclusions.</Paragraph>
  </Section>
class="xml-element"></Paper>