<?xml version="1.0" standalone="yes"?>
<Paper uid="P06-1005">
  <Title>Bootstrapping Path-Based Pronoun Resolution</Title>
  <Section position="4" start_page="0" end_page="33" type="relat">
    <SectionTitle>
2 Related Work
</SectionTitle>
    <Paragraph position="0"> Coreference resolution is generally conducted as a pairwise classi cation task, using various constraints and preferences to determine whether two  expressions corefer. Coreference is typically only allowed between nouns matching in gender and number, and not violating any intrasentential syntactic principles. Constraints can be applied as a preprocessing step to scoring candidates based on distance, grammatical role, etc., with scores developed either manually (Lappin and Leass, 1994), or through a machine-learning algorithm (Kehler et al., 2004). Constraints and preferences have also been applied together as decision nodes on a decision tree (Aone and Bennett, 1995).</Paragraph>
    <Paragraph position="1"> When previous resolution systems handle cases like (1) and (2), where no disagreement or syntactic violation occurs, coreference is therefore determined by the weighting of features or learned decisions of the resolution classi er. Without path coreference knowledge, a resolution process would resolve the pronouns in (1) and (2) the same way. Indeed, coreference resolution research has focused on the importance of the strategy for combining well known constraints and preferences (Mitkov, 1997; Ng and Cardie, 2002), devoting little attention to the development of new features for these dif cult cases. The application of world knowledge to pronoun resolution has been limited to the semantic compatibility between a candidate noun and the pronoun's context (Yang et al., 2005). We show semantic compatibility can be effectively combined with path coreference information in our experiments below.</Paragraph>
    <Paragraph position="2"> Our method for determining path coreference is similar to an algorithm for discovering paraphrases in text (Lin and Pantel, 2001). In that work, the beginning and end nodes in the paths are collected, and two paths are said to be similar (and thus likely paraphrases of each other) if they have similar terminals (i.e. the paths occur with a similar distribution). Our work does not need to store the terminals themselves, only whether they are from the same pronoun group. Different paths are not compared in any way; each path is individually assigned a coreference likelihood.</Paragraph>
  </Section>
class="xml-element"></Paper>