<?xml version="1.0" standalone="yes"?>
<Paper uid="W06-3106">
  <Title>Phrase-Based SMT with Shallow Tree-Phrases</Title>
  <Section position="7" start_page="44" end_page="44" type="evalu">
    <SectionTitle>
4.4 Results
</SectionTitle>
    <Paragraph position="0"> The scores for the 16 slices of the test corpus are reported in Table 2. TP-ENGINE shows slightly better figures for all metrics.</Paragraph>
    <Paragraph position="1"> For each system and for each metric, we had 16 scores (from each of the 16 slices of the test corpus)andwerethereforeabletotestthestatisticalsig- null nicance of the difference between the TP-ENGINE and PP-ENGINE using a Wilcoxon signed-rank test for paired samples. This test showed that the difference observed between the two systems is significant at the 95% probability level for BLEU and significant at the 99% level for WER and SER.</Paragraph>
    <Paragraph position="2">  (+-value range) of the translations produced by the two engines on a test set of 16 disjoint corpora of 500 sentences each. The figures reported are percentages. null On the DEV corpus, we measured that, on average,eachsourcesentenceiscoveredby39TPs(their null source part, naturally), yielding a source coverage of approximately70%. Incontrast, theaveragenumber of covering PPs per sentence is 233.</Paragraph>
  </Section>
class="xml-element"></Paper>