<?xml version="1.0" standalone="yes"?>
<Paper uid="W05-0827">
  <Title>Improving Phrase-Based Statistical Translation by modifying phrase extraction and including several features</Title>
  <Section position="10" start_page="152" end_page="152" type="concl">
    <SectionTitle>
6 Conclusions
</SectionTitle>
    <Paragraph position="0"> We reported a new method to extract longer phrases without increasing the quantity of phrases (less than 0.5%).</Paragraph>
    <Paragraph position="1"> We also reported several features as P(e|f) which in combination with the functions of the source-channel model provides significant improvement. Also, the feature IBM1 in combination with IBM1[?]1 provides improved scores, too.</Paragraph>
    <Paragraph position="2"> Finally, we have optimized the parameters, and we provided the final results which have been presented in the Shared Task: Exploiting Parallel Texts for Statistical Machine Translation (June 30, 2005) in conjunction with ACL 2005 in Ann Arbor, Michigan.</Paragraph>
  </Section>
class="xml-element"></Paper>