<?xml version="1.0" standalone="yes"?>
<Paper uid="P01-1026">
  <Title>Organizing Encyclopedic Knowledge based on the Web and its Application to Question Answering</Title>
  <Section position="7" start_page="4" end_page="4" type="concl">
    <SectionTitle>
6 Conclusion
</SectionTitle>
    <Paragraph position="0"> The World Wide Web has been an unprecedentedly enormous information source, from which a number of language processing methods have been explored to extract, retrieve and discover various types of information. null In this paper, we aimed at generating encyclopedic knowledge, which is valuable for many applications including human usage and natural language understanding. For this purpose, we reformalized an existing Web-based extraction method, and proposed a new statistical organization model to improve the quality of extracted data.</Paragraph>
    <Paragraph position="1"> Given a term for which encyclopedic knowledge (i.e., descriptions) is to be generated, our method sequentially performs a) retrieval of Web pages containing the term, b) extraction of page fragments describing the term, and c) organizing extracted descriptions based on domains (and consequently word senses).</Paragraph>
    <Paragraph position="2"> In addition, we proposed a question answering system, which answers interrogative questions associated with what, by using a Web-based encyclopedia as a knowledge base. For the purpose of evaluation, we used as test inputs technical terms collected from the Class II IT engineers examination, and found that the encyclopedia generated through our method was of operational quality and quantity.</Paragraph>
    <Paragraph position="3"> We also used test questions from the Class II examination, and evaluated the Web-based encyclopedia in terms of question answering. We found that our Web-based encyclopedia improved the system coverage obtained solely with an existing dictionary. In addition, when we used both resources, the performance was further improved.</Paragraph>
    <Paragraph position="4"> Future work would include generating information associated with more complex interrogations, such as ones related to how and why, so as to enhance Web-based natural language understanding.</Paragraph>
  </Section>
class="xml-element"></Paper>