<?xml version="1.0" standalone="yes"?>
<Paper uid="P03-1011">
  <Title>Loosely Tree-Based Alignment for Machine Translation</Title>
  <Section position="7" start_page="4" end_page="4" type="concl">
    <SectionTitle>
6 Conclusion
</SectionTitle>
    <Paragraph position="0"> Our loosely tree-based alignment techniques allow statistical models of machine translation to make use of syntactic information while retaining the flexibility to handle cases of non-isomorphic source and target trees. This is achieved with a clone operation parameterized in such a way that alignment probabilities can be computed with no increase in asymptotic computational complexity.</Paragraph>
    <Paragraph position="1"> We present versions of this technique both for tree-to-string models, making use of parse trees for one of the two languages, and tree-to-tree models, which make use of parallel parse trees. Results in terms of alignment error rate indicate that the clone operation results in better alignments in both cases.</Paragraph>
    <Paragraph position="2"> On our Korean-English corpus, we found roughly equivalent performance for the unstructured IBM models, and the both the tree-to-string and tree-to-tree models when using cloning. To our knowledge these are the first results in the literature for tree-to-tree statistical alignment. While we did not see a benefit in alignment error from using syntactic trees in both languages, there is a significant practical benefit in computational efficiency. We remain hopeful that two trees can provide more information than one, and feel that extensions to the &amp;quot;loosely&amp;quot; tree-based approach are likely to demonstrate this using larger corpora.</Paragraph>
    <Paragraph position="3"> Another important question we plan to pursue is the degree to which these results will be borne out with larger corpora, and how the models may be refined as more training data is available. As one example, our tree representation is unlexicalized, but we expect conditioning the model on more lexical information to improve results, whether this is done by percolating lexical heads through the existing trees or by switching to a strict dependency representation. null</Paragraph>
  </Section>
class="xml-element"></Paper>