<?xml version="1.0" standalone="yes"?>
<Paper uid="W99-0625">
  <Title>Normalized? Yes Yes Yes No Yes No Yes Yes Yes Yes Yes Yes</Title>
  <Section position="8" start_page="209" end_page="210" type="concl">
    <SectionTitle>
7 Conclusion and Future Work
</SectionTitle>
    <Paragraph position="0"> We have presented a new method to detect similarity between small textual units, which combines primitive and composite features using machine learning. We validated our similarity definition using human judges, applied  our method to a substantial number of paragraph pairs from news articles, and compared results to baseline and standard information retrieval techniques. Our results indicate that our method outperforms the standard techniques for detecting similarity, and the system has been successfully integrated into a larger multiple-document summarization system \[McKeown et al. 1999\].</Paragraph>
    <Paragraph position="1"> We are currently working on incorporating a clustering algorithm in order to give as output a set of textual units which are mutually similar rather than just pairwise similar. Future work includes testing on textual units of different size, comparing with additional techniques proposed for document similarity in the information retrieval and computational linguistics literature, and extending the feature set to incorporate other types of linguistic information in the statistical learning method.</Paragraph>
  </Section>
class="xml-element"></Paper>