<?xml version="1.0" standalone="yes"?>
<Paper uid="W04-3219">
  <Title>Monolingual Machine Translation for Paraphrase Generation</Title>
  <Section position="8" start_page="8" end_page="8" type="concl">
    <SectionTitle>
7 Conclusions
</SectionTitle>
    <Paragraph position="0"> We presented a novel approach to the problem of generating sentence-level paraphrases in a broad semantic domain. We accomplished this by using methods from the field of SMT, which is oriented toward learning and generating exactly the sorts of alternations encountered in monolingual paraphrase. We showed that this approach can be used to generate paraphrases that are preferred by humans to sentence-level paraphrases produced by other techniques. While the alternations our system produces are currently limited in character, the field of SMT offers a host of possible enhancements--including reordering models--affording a natural path for future improvements.</Paragraph>
    <Paragraph position="1"> A second important contribution of this work is a method for building and tracking the quality of large, alignable monolingual corpora from structured news data on the Web. In the past, the lack of such a data source has hampered paraphrase research; our approach removes this obstacle.</Paragraph>
  </Section>
class="xml-element"></Paper>