<?xml version="1.0" standalone="yes"?>
<Paper uid="C04-1179">
  <Title>FrameNet-based Semantic Parsing using Maximum Entropy Models</Title>
  <Section position="7" start_page="5" end_page="5" type="concl">
    <SectionTitle>
6 Conclusion
</SectionTitle>
    <Paragraph position="0"> We describe a pipeline framework to analyze sentences into frame elements and semantic roles based on the FrameNet corpus. The process includes four steps: sentence segmentation, FE identification, role classification, and final re-ranking of the n-best outputs.</Paragraph>
    <Paragraph position="1"> In future work, we will investigate ways to reduce the gap between the five-best output performance and the single best output. More features should be extracted to improve re-ranking accuracy. Although the segmentation improves the performance, since the final output is dominated by the initial segmentation, we will explore a smart segmentation method, possibly one not even limited to constituents.</Paragraph>
    <Paragraph position="2"> In addition to the provided syntactic features, we will apply semantic features using ontology.</Paragraph>
    <Paragraph position="3"> Finally, the challenge is to apply this type of work to new predicates, ones not yet treated in FrameNet. We are searching for methods to achieve this.</Paragraph>
  </Section>
class="xml-element"></Paper>