<?xml version="1.0" standalone="yes"?>
<Paper uid="C86-1089">
  <Title>Learning the Space of Word Meanings for Information Retrieval Systems</Title>
  <Section position="7" start_page="378" end_page="378" type="concl">
    <SectionTitle>
6. Conclusion
</SectionTitle>
    <Paragraph position="0"> We proposed to represent the meanings of words in space. We gave a learning mechanism for tile space. We proved tile effectiveness of the proposed methods in an information retrieval system.</Paragraph>
    <Paragraph position="1"> One of the reasons we chose the field of Japanese literature was that there was strong demmld from th.e researchers in that field to make a useful system; systems based on conventional key words did riot work welt for the field of literature.</Paragraph>
    <Paragraph position="2"> Another reason is that the field offers good examples of entities for which universal symbolic representation is difficult.</Paragraph>
    <Paragraph position="3"> To apply our method to other fields, we must solve only two problems. One problem is to select symbols to be put ill tile initial space. In literature, works and writers played tile role. Ti~e other problem is to determine criteria for space configuration. We used a user's satisfaction wittl the answers from the information retrieval system as tile criterion. If these two problems are solved, our method can be applied to any domain. &amp;quot;l'hcsc problems do uot seem so difficult.</Paragraph>
    <Paragraph position="4"> From the viewpoint of artificial intelligence, the.</Paragraph>
    <Paragraph position="5"> semantic space Rives a basis for studies on abduetion an.d analogy. The diseovery of blank space snr.rounded by symbols carl lead to discovering of new ideas by mac\]tines. Siuce we can measure semantic similarity directly by distance in semantic space, wc can make analogical reasoning based oil that similar-. ity. For the same reason, from the linguistic point of view, the semantic spac, e e~l be useful for under..</Paragraph>
    <Paragraph position="6"> standing metaphorical expression such as &amp;quot;Hofs..</Paragraph>
    <Paragraph position="7"> tadter is Prolog between procedural semantics and declarative semantics&amp;quot;.</Paragraph>
  </Section>
class="xml-element"></Paper>