<?xml version="1.0" standalone="yes"?>
<Paper uid="N04-4012">
  <Title>UI on the Fly: Generating a Multimodal User Interface</Title>
  <Section position="6" start_page="2" end_page="2" type="evalu">
    <SectionTitle>
6 Conclusion
</SectionTitle>
    <Paragraph position="0"> We have demonstrated a formalism that generates coherent multimodal user interfaces, as well its application in a small-screen email client. As the generation algorithm makes use of both hard constraints and scalar scores, it caters for adaptability. We have proven its functionality and efficiency in a series of examples in the context of a dialogue system, where content is generated in real-time for various usage situations and different devices.</Paragraph>
    <Paragraph position="1"> Further evaluation will show whether the fitness function can accurately mirror user satisfaction with a given output variant and whether our form of adaptivity is actually an advantage to users on the go. Without a gold standard for a generation system for dynamic multimodal user interfaces to qualitatively compare against, controlled user trials will allow us to evaluate the usability of the interfaces we have created. Task completion times, user frustration levels, and user satisfaction can then be used to evaluate the success of this model of multimodal interactions.</Paragraph>
    <Paragraph position="2"> The underlying formalism is intended to be used in creating, using the MUG Workbench, any multimodal system that can be constructed compositionally, using natural language and other auditory and visual components.</Paragraph>
    <Paragraph position="3"> As possible examples for future applications, we see a multimodal interface that allows mobile users or users with sensory impairments to traverse information-rich social networks, and a kiosk for multimodal, multilingual access to public transportation options.</Paragraph>
  </Section>
class="xml-element"></Paper>