<?xml version="1.0" standalone="yes"?>
<Paper uid="W04-2413">
  <Title>Semantic Role Labelling With Chunk Sequences</Title>
  <Section position="2" start_page="0" end_page="0" type="intro">
    <SectionTitle>
1 Introduction
</SectionTitle>
    <Paragraph position="0"> This paper describes a statistical approach to semantic role labelling addressing the CoNLL shared task 2004, which is based on the the current release of the English PropBank data (Kingsbury et al., 2002). For further details of the task, see (Carreras and Marquez, 2004).</Paragraph>
    <Paragraph position="1"> We address the main challenge of the task, the absence of deep syntactic information, with three main ideas: a0 Proper constituents being unavailable, we use chunk sequences as instances for classification.</Paragraph>
    <Paragraph position="2"> a0 The classification is performed by a maximum entropy model, which can integrate features from heterogeneous data sources.</Paragraph>
    <Paragraph position="3"> a0 We model the fit between verb and argument candidate by clusters induced with EM on the training data, which we use as features during classification.</Paragraph>
    <Paragraph position="4"> Sections 2 through 4 describe the systems' architecture. First, we compute chunk sequences for all sentences (Sec. 2). Then, we classify these sequences with maximum entropy models (Sec. 3). Finally, we determine the most probable chain of sequences covering the whole sentence (Sec. 4). Section 5 discusses the impact of different parameters and gives final results.</Paragraph>
  </Section>
class="xml-element"></Paper>