<?xml version="1.0" standalone="yes"?>
<Paper uid="W05-0635">
  <Title>Semantic Role Labeling Using Complete Syntactic Analysis</Title>
  <Section position="6" start_page="223" end_page="223" type="concl">
    <SectionTitle>
4 Conclusions
</SectionTitle>
    <Paragraph position="0"> This paper describes a semantic role labeling system constructed on top of the complete syntactic analysis of text. We model semantic arguments that map into exactly one syntactic phrase (about 90% of all semantic arguments in the development set) using a rich set of lexical, syntactic, and semantic attributes. We trained AdaBoost one-versus-all classifiers for the 24 most common argument types. Arguments that map to more than one syntactic constituent are expanded with a simple heuristic in a post-processing step.</Paragraph>
    <Paragraph position="1"> Our results surpass with almost 6% the results obtained by best SRL system that used partial syntax in the CoNLL 2004 shared task evaluation. Although the two evaluations are not directly comparable due to differences in training set size, the current results are encouraging given the simplicity of our proposed system.</Paragraph>
  </Section>
class="xml-element"></Paper>