<?xml version="1.0" standalone="yes"?>
<Paper uid="A92-1033">
  <Title>Overview of Natural Language Processing of Captions for Retrieving Multimedia Data</Title>
  <Section position="4" start_page="231" end_page="231" type="metho">
    <SectionTitle>
3 Implementation Status
</SectionTitle>
    <Paragraph position="0"> The majority of the system is written in Quintus Prolog.</Paragraph>
    <Paragraph position="1"> with the type hierarchy being developed using the Elsa-Lap object-oriented Prolog tool. The system runs on Sun Sparcstations and was designed using a client-server relationship; the user search environment and key creation interface form the two clients and a server process handles the parsing of the natural language, generation of the keys, and the matching. The lexicon has over 1000 lexical items and the type hierarchy has over 200 classes. Further implementation and methodology details can be found in Guglielmo and Rowe (1991).</Paragraph>
  </Section>
  <Section position="5" start_page="231" end_page="231" type="metho">
    <SectionTitle>
4 Future Research
</SectionTitle>
    <Paragraph position="0"> The present system handles individual captions that describe an individual photograph. Future work will investigate supercaptions. For example, supercaptions that are used to represent all captions from the same chapter of a book or a supercaption that is used to represent all captions that pertain to a combat plan. All of the member captions share something in common, and the intersection of this common information forms the supercaption.</Paragraph>
  </Section>
class="xml-element"></Paper>