<?xml version="1.0" standalone="yes"?>
<Paper uid="I05-2009">
  <Title>Analysis and modeling of manual summarization of Japanese broadcast news</Title>
  <Section position="14" start_page="11221" end_page="11221" type="concl">
    <SectionTitle>
6 Conclusions
</SectionTitle>
    <Paragraph position="0"> We have described the manual summary process of NHK's broadcast news and experiments on automatic text alignment between news summaries and the original news together  The present summary length constraint is 105 characters. Meanwhile, the average length of the first sentence (typically the lead) of a news text is 94.5 as is shown in table 1. with the word position matching. Through a statistical analysis of the results and interviews with abstractors, we found that the abstractors summarize news by taking advantage of its structure. Based on this observation, we proposed a summarization model that consists of a reading and editing phase. We are now designing an automatic or semi automatic summarization system employing the model.</Paragraph>
  </Section>
class="xml-element"></Paper>