<?xml version="1.0" standalone="yes"?>
<Paper uid="E06-2015">
  <Title>Semantic Role Labeling for Coreference Resolution</Title>
  <Section position="6" start_page="145" end_page="145" type="concl">
    <SectionTitle>
4 Conclusion
</SectionTitle>
    <Paragraph position="0"> In this paper we have investigated the effects of using semantic role information within a machine learning based coreference resolution system. Empirical results show that coreference resolution can benefit from SRL. The analysis of the relevance of features, which had not been previously addressed, indicates that incorporating semantic information as shallow event descriptions improves the performance of the classifier. The generated model is able to learn selection preferences in cases where surface morpho-syntactic features do not suffice, i.e. pronoun resolution.</Paragraph>
    <Paragraph position="1"> We speculate that this contrasts with the disappointingfindingsofKehleretal. (2004)sinceSRL provides a more fine grained level of information when compared to predicate argument statistics.</Paragraph>
    <Paragraph position="2"> As it models the semantic relationship that a syntacticconstituenthaswithapredicate, itcarriesindirectly syntactic preference information. In addition, when used as a feature it allows the classifier to infer semantic role co-occurrence, thus inducing deep representations of the predicate argument relations for learning in coreferential contexts.</Paragraph>
    <Paragraph position="3"> Acknowledgements: This work has been funded  bytheKlausTschiraFoundation,Heidelberg,Germany. The first author has been supported by a KTF grant (09.003.2004).</Paragraph>
  </Section>
class="xml-element"></Paper>