<?xml version="1.0" standalone="yes"?>
<Paper uid="E06-1036">
  <Title>Recognizing Textual Parallelisms with edit distance and similarity degree</Title>
  <Section position="7" start_page="287" end_page="287" type="concl">
    <SectionTitle>
6 Conclusion
</SectionTitle>
    <Paragraph position="0"> Textual parallelism plays an important role among discourse features when detecting discourse structures. Sofar, onlyoccurrences ofthisphenomenon have been treated individually and often in an ad-hoc manner. Our contribution is a unifying framework which can be used for automatic processing with much less specific knowledge than dedicated techniques.</Paragraph>
    <Paragraph position="1"> In addition, we discussed and evaluated several methods to retrieve them generically. We showed that simple methods such as (Wagner and Fischer, 1974) can compete with more complex approaches, such as our degree of similarity and the 4Compared to entailment, the parallelism relation is bi-directional and not restricted to semantic similarities. (Zhang and Shasha, 1989)'s algorithm.</Paragraph>
    <Paragraph position="2"> Among future works, it seems that variations such as the editing cost of transformation for edit distance methods and the weight of parallel units (depending their semantic and syntactic characteristics) can be implemented to enhance performances. Combining methods also seems an interesting track to follow.</Paragraph>
  </Section>
class="xml-element"></Paper>