<?xml version="1.0" standalone="yes"?>
<Paper uid="W05-1204">
  <Title>Training Data Modification for SMT Considering Groups of Synonymous Sentences</Title>
  <Section position="7" start_page="23" end_page="23" type="concl">
    <SectionTitle>
7 Conclusion
</SectionTitle>
    <Paragraph position="0"> This paper described the modification of a training set based on a synonymous sentence group for a statistical machine translation system in order to attain better performance. In an EJ/JE system, we confirmed a positive effect by replacing the input-side language. Because the Chinese data was specific in our modification, we observed an inconclusive result for the modification in the CJ/JC system based on the synonymous sentence group with a JC language pair. However, there was still some effect on the characteristics of the training corpus. In this paper, the modifications of the training set are based on the synonymous sentence group, and we replace the sentence with rough processing. If we paraphrased the training set and controlled the phrase pair, we could achieve better performance with the same training set.</Paragraph>
  </Section>
class="xml-element"></Paper>