<?xml version="1.0" standalone="yes"?>
<Paper uid="W04-0857">
  <Title>Generative Models for Semantic Role Labeling</Title>
  <Section position="8" start_page="2" end_page="2" type="concl">
    <SectionTitle>
7 Conclusions
</SectionTitle>
    <Paragraph position="0"> In conclusion, our generative model performs robustly on the easy version of the SENSEVAL-3 role labeling task. The combination of our constituent classifier with the role labeling has more room for improvement, but performed reasonably well considering the difficulties of the task and the sparse feature set that we incorporated into our generative model. Alternative sentence chunking models for semantic analysis, and the extension of our generative models, should lead to future improvements.</Paragraph>
    <Paragraph position="1"> The key advantage of our approach is the treatment of a sentence's roles as a sequence. This allows the model to consider relationships between roles as it semantically analyzes a sentence.</Paragraph>
  </Section>
class="xml-element"></Paper>