<?xml version="1.0" standalone="yes"?>
<Paper uid="C04-1030">
  <Title>Reordering Constraints for Phrase-Based Statistical Machine Translation</Title>
  <Section position="6" start_page="0" end_page="0" type="relat">
    <SectionTitle>
5 Related Work
</SectionTitle>
    <Paragraph position="0"> Recently, phrase-based translation approaches became more and more popular. Marcu and Wong (2002) present a joint probability model for phrase-based translation. In (Koehn et  for the SLDB task (330 sentences). Sentence lengths: short: &lt; 10 words, long: , 10 words; times in milliseconds per sentence.</Paragraph>
    <Paragraph position="1">  al., 2003), various aspects of phrase-based systems are compared, e.g. the phrase extraction method, the underlying word alignment model, or the maximum phrase length. In (Vogel, 2003), a phrase-based system is used that allows reordering within a window of up to three words. Improvements for a Chinese-English task are reported compared to a monotone search.</Paragraph>
    <Paragraph position="2"> The ITG constraints were introduced in (Wu, 1995). The applications were, for instance, the segmentation of Chinese character sequences into Chinese words and the bracketing of the source sentence into sub-sentential chunks. Investigations on the IBM constraints (Berger et al., 1996) for single-word based statistical machine translation can be found e.g. in (Tillmann and Ney, 2003). A comparison of the ITG constraints and the IBM constraints for single-word based models can be found in (Zens and Ney, 2003). In this work, we investigated these reordering constraints for phrase-based statistical machine translation.</Paragraph>
  </Section>
class="xml-element"></Paper>