<?xml version="1.0" standalone="yes"?>
<Paper uid="N06-3003">
  <Title>Can the Internet help improve Machine Translation?</Title>
  <Section position="8" start_page="221" end_page="221" type="evalu">
    <SectionTitle>
7 Evaluation
</SectionTitle>
    <Paragraph position="0"> We plan to evaluate the RR module on its ability to improve coverage and overall translation quality.</Paragraph>
    <Paragraph position="1"> This requires identifying sensible evaluation metrics. Initial experiments have shown that both BLEU [Papineni et al., 2001] and METEOR [Lavie et al., 2004] can automatically distinguish between raw MT output and corrected MT output, even for a small set of sentences. In addition to the presence of the corrected translation in the lattice produced by the refined system, our evaluation metrics will also need to take into account whether the incorrect translation is now prevented from being generated and whether the lattice of alternative translations increased or decreased. A decrease of lattice size would mean that the refinement also made the grammar tighter, which is the desired effect.</Paragraph>
  </Section>
class="xml-element"></Paper>