<?xml version="1.0" standalone="yes"?>
<Paper uid="C04-1030">
  <Title>Reordering Constraints for Phrase-Based Statistical Machine Translation</Title>
  <Section position="7" start_page="0" end_page="0" type="concl">
    <SectionTitle>
6 Conclusions
</SectionTitle>
    <Paragraph position="0"> We have presented different reordering constraints for phrase-based statistical machine translation, namely the IBM constraints and the ITG constraints, as well as efficient dynamic programming algorithms. Translation results were reported for two Japanese-English translation tasks. Both type of re-ordering constraints resulted in improvements compared to a monotone search. Restricting the reorderings according to the IBM constraints resulted already in a translation quality similar to an unconstrained search. The translation results with the ITG constraints even outperformed the unconstrained search consistently on all error criteria. The improvements have been found statistically significant. The ITG constraints showed the best performance on both tasks. Therefore we plan to further improve this method. Currently, the probability model for the ITG constraints is very simple. More sophisticated models, such as phrase dependent inversion probabilities, might be promising.</Paragraph>
  </Section>
class="xml-element"></Paper>