<?xml version="1.0" standalone="yes"?>
<Paper uid="I05-2016">
  <Title>Controlling Animated Agents in Natural Language</Title>
  <Section position="2" start_page="0" end_page="0" type="intro">
    <SectionTitle>
1 Introduction
</SectionTitle>
    <Paragraph position="0"> SHRDLU (?) can be considered as the most important natural language understanding system.</Paragraph>
    <Paragraph position="1"> Although SHRDLU was not &amp;quot;embodied&amp;quot;, having had only a small stick to manipulate objects, it certainly had several features that a conversational agent is supposed to have. It had a great potential, and it was very promising for future research on natural language understanding.</Paragraph>
    <Paragraph position="2"> Recently better technologies have become available in speech recognition and natural language processing. Major breakthroughs in the area of computer graphics have enabled us to generate complex, yet realistic 3-D animated agents or embodied life-like agents in a virtual environment. Researchers are now in a good position to go beyond SHRDLU by combining these technologies (?). This paper presents a conversational animated agent system, K3.</Paragraph>
    <Paragraph position="3"> Since all the actions carried out by an agent of the K3 system are visible, we can evaluate the performance of the system by observing its animation. Visualizing the agents' actions yields many interesting issues from a cognitive science point of view; more complex processes are involved than those found in most conventional natural language understanding systems.</Paragraph>
    <Paragraph position="4"> After sketching out the overview of the K3 system in section 2, Two distinctive features of K3 are discussed in section 3, and 4. Finally, section 5 concludes the paper and looks at future research agenda.</Paragraph>
  </Section>
class="xml-element"></Paper>