<?xml version="1.0" standalone="yes"?>
<Paper uid="W03-1202">
  <Title>Using Thematic Information in Statistical Headline Generation</Title>
  <Section position="7" start_page="0" end_page="0" type="concl">
    <SectionTitle>
7 Conclusion
</SectionTitle>
    <Paragraph position="0"> Combining both the SVD probability and Conditional probability marginally improves recall, lending support to the intuition that thematic information may help generate better single sentence summaries. However, there are still many unanswered questions. In future work, we intend to investigate these techniques in a domain other than news text so that we can draw conclusions as to how well these strategies generalise to other genres. We also intend to conduct user evaluations to gauge the quality of the generated summaries for both the Conditional and the SVD strategies. Indeed, a user-based evaluation would be extremely helpful in determining if the thematic information provided by the SVD strategy does help improve the veracity of the generated summaries.</Paragraph>
  </Section>
class="xml-element"></Paper>