<?xml version="1.0" standalone="yes"?>
<Paper uid="W01-0813">
  <Title>Applying Natural Language Generation to Indicative Summarization</Title>
  <Section position="8" start_page="0" end_page="0" type="concl">
    <SectionTitle>
8 Conclusion
</SectionTitle>
    <Paragraph position="0"> We have presented a model for indicative multidocument summarization based on natural language generation. In our model, summary content is based on document features describing topic and structure instead of extracted text. Given these features, a generation model uses a text plan, derived from analysis of naturally occurring indicative summaries plus guidelines for summarization, to guide the system in describing document topics as typical, rare, intricate, or relevant to the user query. We showed how the topicality document feature can be derived from the set of input documents and represented as a topic tree for each document along with a merged composite topic for all documents in the collection against which prototypicality and query relevance can be computed. Our ongoing work is examining how to automatically learn the text plans along with the tactics needed to realize each piece of the instantiated plan as a sentence.</Paragraph>
  </Section>
class="xml-element"></Paper>