<?xml version="1.0" standalone="yes"?>
<Paper uid="W01-1406">
  <Title>A best-first alignment algorithm for automatic extraction of transfer mappings from bilingual corpora</Title>
  <Section position="5" start_page="0" end_page="0" type="concl">
    <SectionTitle>
6 Conclusions and Future Work
</SectionTitle>
    <Paragraph position="0"> We proposed an algorithm for automatically acquiring high-quality transfer mappings from sentence-aligned bilingual corpora using an alignment grammar and a best-first strategy.</Paragraph>
    <Paragraph position="1"> We reported the results of applying the algorithm to a substantially larger training corpus than that used in previously reported work on learning transfer mappings from parsed data.</Paragraph>
    <Paragraph position="2"> We showed that this approach produces transfer mappings that result in translation quality comparable to a commercial MT system for this domain.</Paragraph>
    <Paragraph position="3"> We also showed that a best-first, alignmentgrammar based approach produced better results than a bottom-up approach, and that retaining context in the acquired transfer mappings is essential to translation quality.</Paragraph>
    <Paragraph position="4"> We currently rely on a priori linguistic heuristics to try to provide the right context for each transfer mapping. In future work, we plan to use machine-learning techniques to determine the extent of the context that optimally disambiguates between conflicting mappings.</Paragraph>
  </Section>
class="xml-element"></Paper>