<?xml version="1.0" standalone="yes"?>
<Paper uid="P04-3014">
  <Title>Improving Bitext Word Alignments via Syntax-based Reordering of English</Title>
  <Section position="6" start_page="35" end_page="35" type="evalu">
    <SectionTitle>
5 Results
</SectionTitle>
    <Paragraph position="0"> Figures 4, 5, 6 and 7 show learning curves for systems trained on parallel sentences with and without the Englishprime transforms. Table 2 provides further detail, and also shows the performance of systems trained without any bitext, but only with access to a bilingual translation lexicon. Our system achieves consistent, substantial performance improvement under all situations for English-Hindi and English-Korean language pairs, which exhibit longer distance SOV-SVO syntactic divergence.</Paragraph>
    <Paragraph position="1"> For English-Romanian and English-Chinese, neither significant improvement nor degradation is seen, but these are language pairs with quite similar sentential word order to English, and hence have less opportunity to benefit from our syntactic transformations.</Paragraph>
  </Section>
class="xml-element"></Paper>