<?xml version="1.0" standalone="yes"?>
<Paper uid="W04-2804">
  <Title>Ends-based Dialogue Processinga0</Title>
  <Section position="2" start_page="0" end_page="0" type="intro">
    <SectionTitle>
1 Introduction
</SectionTitle>
    <Paragraph position="0"> In the last couple of years our group at DFKI in Saarbr&amp;quot;ucken has been involved in a number of projects aiming at interfacing different devices in an intelligent way. The main goal of these projects has been to build functioning robust systems with which it is natural to communicate (not only for some few examples phrases).</Paragraph>
    <Paragraph position="1"> During the projects we have developed a dialogue toolbox consisting of a number of modules. By combining these modules in different ways we are able to realize a number of different types of dialogues, e. g., information seeking/browsing, device control, multi/cross-application and agent-mediated interactions for a number of (diverse) applications and systems. The full-blown combination of all modules form our dialogue backbone capable of engaging in multimodal man-machine communication.</Paragraph>
    <Paragraph position="2"> In this paper, we discuss some of the design decisions taken along the road as well as lessons learned during the projects. Based on our experiences, we argue that ends-based processing is vital to the success of our approach. We strive for a balance between complex theories and pragmatic decisions. Of secondary interest is the implementation of theories capable of processing linguistically exotic phenomena in favor of ends-based processing in all modules of the toolbox. Hence it is more important to reach the representation rather than how we get there.</Paragraph>
    <Paragraph position="3"> An ontology is often - as we understand it - a good ends-based representation but we can do without it. In the MIAMM project (see section 2) we use no ontology a3The research presented here is funded by the German Ministry of Research and Technology under grant 01 IL 905, the European Union under the grants IST-2000-29487 and IST-200132311 and IDS-Scheer AG.</Paragraph>
    <Paragraph position="4"> but instead an event based representation. Whatever representation we do choose, we would like to stress the importance of a consequent principle-based design of the representation and the fact that the complete backbone uses it. Exactly this guarantees, e. g., the scalability of our approach.</Paragraph>
    <Paragraph position="5"> The paper is organized as follows: the next section provides an overview of projects and systems central to the development of our toolbox. Section 3 describes most of its modules. Before we conclude the paper, we provide a list of claims and lessons learned in section 4.</Paragraph>
  </Section>
class="xml-element"></Paper>