<?xml version="1.0" standalone="yes"?>
<Paper uid="W03-0311">
  <Title>Retrieving Meaning-equivalent Sentences for Example-based Rough Translation</Title>
  <Section position="7" start_page="0" end_page="0" type="concl">
    <SectionTitle>
6 Conclusions
</SectionTitle>
    <Paragraph position="0"> In this paper, we introduced the idea of meaning-equivalent sentences for robust example-based S2ST.</Paragraph>
    <Paragraph position="1"> Meaning-equivalent sentences have the same main meaning as the input despite lacking some unimportant information. Translation of meaning-equivalent sentences corresponds to rough translations, which aim not at exact translation with narrow coverage but at rough translation with wide coverage. For S2ST, we assume that this translation strategy is sufficiently useful.</Paragraph>
    <Paragraph position="2"> Then, we described a method for retrieving meaning-equivalent sentences from an example corpus. Retrieval is based on content words, modality, and tense. This strategy is feasible owing to the restricted domains, often adopted in S2ST, which have relatively small variety in lexicon and meaning. An experiment demonstrated the robustness of our method to input length and the style differences between inputs and the example corpus.</Paragraph>
    <Paragraph position="3"> Most MT systems aim to achieve exact translation, but unfortunately they often output bad or no translation for long conversational speeches. The rough translation proposed in this paper achieves robustness in translation for such inputs. This method compensates for the shortcomings of conventional MT and makes S2ST technology more practical.</Paragraph>
  </Section>
class="xml-element"></Paper>