<?xml version="1.0" standalone="yes"?>
<Paper uid="W96-0201">
  <Title>A Geometric Approach to Mapping Bitext Correspondence</Title>
  <Section position="8" start_page="109000" end_page="109000" type="concl">
    <SectionTitle>
5. Conclusion
</SectionTitle>
    <Paragraph position="0"> The Smooth Injective Map Recognizer (SIMR) has five advantages over previous bitext mapping algorithms. First, it lowers average errors by more than a factor of 4. Second, it avoids very large errors, improving robustness to a level that enables new commercial-quality applications. Third, it does not require large amounts of computer memory to run. Fourth, it accepts non-monotonic segments to account for inversions and word order differences. Fifth, its output can be converted quickly and easily into an accurate sentence alignment. null There are many possible extensions to this work. One interesting observation is that aligned sentences can be used to induce translation lexicons, and translation lexicons are an important information source for bitext mapping and alignment (Kay &amp; RSscheisen 1993, Chen 1993). I plan to explore an interactive loop between SIMR, GSA and my algorithm for inducing translation lexicons (Melamed 1995).</Paragraph>
    <Paragraph position="1"> It would also be interesting to experiment with SIMR and GSA on language pairs that are not as closely related as English and French. The only technique for mapping between more disparate languages that has been rigorously evaluated (Wu 1994) relies on length correlations sprinkled with some lexical information. From this point of view, Wu's technique is similar to the one used by Simard et al. (1992). So, I am eager to see whether the geometric approach will compare as favorably to Wu's results on English and Chinese as it has to Simard et al.'s results on English and French.</Paragraph>
  </Section>
class="xml-element"></Paper>