<?xml version="1.0" standalone="yes"?>
<Paper uid="W99-0204">
  <Title>Automatic Slide Presentation from Semantically Annotated Documents</Title>
  <Section position="2" start_page="0" end_page="0" type="intro">
    <SectionTitle>
1 Introduction
</SectionTitle>
    <Paragraph position="0"> A presentation of information content must be adapted to the context. A problem arises here because of diverse types of contexts mainly due to the audience's idiosyncratic needs, backgrounds, and so forth. Adaptation by learning \[Perkovitz and Etzioni, 1997; 1998\] cannot provide a full solution here, because individual information seekers' profiles and contexts are unpredictable from past experiences. It is essentially necessary to dynamically customize a presentation through interactions with the audience, as human presenters normally do.</Paragraph>
    <Paragraph position="1"> In the present paper we discuss how to automatically generate slide shows from semantically annotated documents, in such a way that the presentation can be dynamically adapted to the audience. The reported presentation system detects important topics in the input document and composes a slide for each topic by extracting and paraphrasing relevant Sentences. This whole process takes into consideration not only the semantic structure of the given document but also interactions with the audience. So the slide show can be dynamically customized by reflecting requests and queries from the audience during the presentation.</Paragraph>
    <Paragraph position="2"> Each slide is typically an itemized summary of a topic in the original document. Generating such slides and coordinating them to meet the audience's needs involves a lot more drastic reformation of the original document than mere extraction of sentences in traditional summarization, so that accurate semantic structure of the document is necessary. We hence assume that the input documents come with GDA (Global Document Annotation) tags \[Hasida, 1997; Nagao and Hasida, 1998\] embedded. The GDA tagset is an XML (eXtensible Markup Language) tagset which allows machines to automatically infer the semantic structures (including pragmatic structures) underlying the raw documents.</Paragraph>
    <Paragraph position="3"> Under the current state of the art, GDA-tagging can be only semiautomatic and calls for manual correction.</Paragraph>
    <Paragraph position="4"> The cost involved here pays, because an annotated document is a generic form of information content from which to compose diverse types of presentations, potentially involving summarization, narration, visualization, translation, information retrieval, information extraction, and so forth. The slide presentation system reported below addresses a core technology in this broad setting. In the rest of the paper, we first outline the GDA tagset, and discuss how to extract topics from the input document and to generate slides for them by exploiting the tags.</Paragraph>
  </Section>
class="xml-element"></Paper>