<?xml version="1.0" standalone="yes"?>
<Paper uid="N04-4003">
  <Title>Example-based Rescoring of Statistical Machine Translation Output</Title>
  <Section position="7" start_page="0" end_page="0" type="concl">
    <SectionTitle>
6 Conclusion
</SectionTitle>
    <Paragraph position="0"> In this paper, we proposed an example-based method for selecting translation candidates generated by a statistical decoder. It utilizes translation examples that are similar to the source sentence as the input and validates the decoder output against its seed sentences in order to identify defective translations. The revised scoring scheme achieved a translation accuracy of 92.7%, an improvement of 11.9% over the baseline system.</Paragraph>
    <Paragraph position="1"> So far, we treated the statistical decoder as a blackbox. However, further investigations will have to separate modeling errors and search errors during decoding and compare our findings to advanced statistical modeling approaches (phrase-based) and other search strategies. Future work will also focus on the integration of the proposed rescoring formula in the decoding process.</Paragraph>
  </Section>
class="xml-element"></Paper>