<?xml version="1.0" standalone="yes"?>
<Paper uid="W05-0822">
  <Title>PORTAGE: A Phrase-based Machine Translation System</Title>
  <Section position="6" start_page="131" end_page="131" type="concl">
    <SectionTitle>
4 Conclusion
</SectionTitle>
    <Paragraph position="0"> We have reported on our participation in the shared task of the ACL 2005 Workshop on Building and Using Parallel Texts, conducting evaluations of Portage, our statistical machine translation system, on all four language pairs. Our best BLEU scores for the French-, Finnish-, German-, and Spanish-English at this stage were 29.5, 20.95, 23.21 and 29.08, respectively. In total, eleven teams took part at the shared task and most of them submitted results for all pairs of languages. Our results distinguished the NRC team at the third, second, third and fourth ranks with slight differences with the first ranked participants.</Paragraph>
    <Paragraph position="1"> A major goal of this work was to evaluate Portage at its first stage of implementation on different pairs of languages. This evaluation has served to identify some problems with our system in the areas of weight optimization and number and date rules.</Paragraph>
    <Paragraph position="2"> It has also indicated the limits of using out-of-domain corpora, and the difficulty of morphologically complex languages like Finnish.</Paragraph>
    <Paragraph position="3"> Current and planned future work includes the exploitation of comparable corpora for statistica machine transl knowledge, and better features for nbest rescoring.</Paragraph>
  </Section>
class="xml-element"></Paper>