<?xml version="1.0" standalone="yes"?>
<Paper uid="W05-0627">
  <Title>Semantic Role Lableing System using Maximum Entropy Classi er [?]</Title>
  <Section position="7" start_page="191" end_page="191" type="concl">
    <SectionTitle>
4 Conclusions
</SectionTitle>
    <Paragraph position="0"> We have described a maximum entropy classi er is our semantic role labeling system, which takes syntactic constituents as the labeling units. The fast training speed of the maximum entropy classi er allows us just use one stage of arguments identi cation and classi cation to build the system.</Paragraph>
    <Paragraph position="1"> Some useful features and their combinations are evaluated. Only the constituents with the largest probability among embedding ones are kept. After predicting all arguments which have matching constituents in full parsing trees, a simple rule-based post-processing is applied to correct the arguments which have no matching constituents. The constituent-based method depends much on the syntactic parsing performance. The comparison between WSJ and Brown test sets results fully demonstrates the point of view.</Paragraph>
  </Section>
class="xml-element"></Paper>