<?xml version="1.0" standalone="yes"?>
<Paper uid="I05-5008">
  <Title>Automatic generation of paraphrases to be used as translation references in objective evaluation measures of machine translation</Title>
  <Section position="4" start_page="0" end_page="0" type="intro">
    <SectionTitle>
2 Motivation
</SectionTitle>
    <Paragraph position="0"> Paraphrases are an important element in the evaluation of many natural language processing tasks.</Paragraph>
    <Paragraph position="1"> Specifically, in the automatic evaluation of machine translation systems, the quality of translation candidates is judged against reference translations that are paraphrases in the target language.</Paragraph>
    <Paragraph position="2"> Automatic measures like BLEU (PAPINENI et al., 2001) or NIST (DODDINGTON, 2002) do so by counting sequences of words in such paraphrases.</Paragraph>
    <Paragraph position="3"> It is expected that such reference sets contain synonymous sentences (i.e., paraphrases) that explicit possible lexical and syntactical variations in order to cope with translation variations in terms and structures (BABYCH and HARTLEY, 2004).</Paragraph>
    <Paragraph position="4"> In order to produce such reference sets, we propose a method to generate paraphrases from a seed sentence where lexical and syntactical variations are handled by the use of commutations as captured by proportional analogies whereas N-sequences are used to enforce fluency of expression and adequacy of meaning.</Paragraph>
  </Section>
class="xml-element"></Paper>