<?xml version="1.0" standalone="yes"?>
<Paper uid="P03-1040">
  <Title>Feature-Rich Statistical Translation of Noun Phrases</Title>
  <Section position="6" start_page="0" end_page="0" type="evalu">
    <SectionTitle>
5 Results
</SectionTitle>
    <Paragraph position="0"> As described in Section 3.1, we evaluate the performance of our NP/PP translation subsystem on a blind test set of 1362 NP/PPs extracted from 534 sentences. The contributions of different components of our system are displayed in Table 3.</Paragraph>
    <Paragraph position="1"> Starting from the IBM Model 4 baseline, we achieve gains using our phrase-based translation model (+5.5%), applying compound splitting to  special modeling and additional features: Correct NP/PPs and BLEU score for overall sentence translation null training and test data (+2.8%), re-estimating the weights for the system components using the maximum entropy reranking frame-work (+1.5%), adding web count features (+1.7%) and syntactic features (+0.8%). Overall we achieve an improvement of 12.3% over the baseline. Improvements of 2.5% are statistically significant given the size of our test corpus.</Paragraph>
    <Paragraph position="2"> Table 3 also provides scores for overall sentence translation quality. The chosen NP/PP translations are integrated into a general IBM Model 4 system that translates whole sentences. Performance is measured by the BLEU score, which measures similarity to a reference translation. As reference translation we used the English side of the parallel corpus. The BLEU scores track the improvements of our components, with an overall gain of 0.027.</Paragraph>
  </Section>
class="xml-element"></Paper>