<?xml version="1.0" standalone="yes"?>
<Paper uid="P04-3033">
  <Title>MATCHKiosk: A Multimodal Interactive City Guide</Title>
  <Section position="2" start_page="0" end_page="0" type="intro">
    <SectionTitle>
1 Introduction
</SectionTitle>
    <Paragraph position="0"> Since the introduction of automated teller machines in the late 1970s, public kiosks have been introduced to provide users with automated access to a broad range of information, assistance, and services. These include self check-in at airports, ticket machines in railway and bus stations, directions and maps in car rental offices, interactive tourist and visitor guides in tourist offices and museums, and more recently, automated check-out in retail stores. The majority of these systems provide a rigid structured graphical interface and user input by only touch or keypad, and as a result can only support a small number of simple tasks. As automated kiosks become more commonplace and have to support more complex tasks for a broader community of users, they will need to provide a more flexible and compelling user interface.</Paragraph>
    <Paragraph position="1"> One major motivation for developing multimodal interfaces for mobile devices is the lack of a keyboard or mouse (Oviatt and Cohen, 2000; Johnston and Bangalore, 2000). This limitation is also true of many different kinds of public information kiosks where security, hygiene, or space concerns make a physical keyboard or mouse impractical. Also, mobile users interacting with kiosks are often encumbered with briefcases, phones, or other equipment, leaving only one hand free for interaction. Kiosks often provide a touchscreen for input, opening up the possibility of an onscreen keyboard, but these can be awkward to use and occupy a considerable amount of screen real estate, generally leading to a more moded and cumbersome graphical interface.</Paragraph>
    <Paragraph position="2"> A number of experimental systems have investigated adding speech input to interactive graphical kiosks (Raisamo, 1998; Gustafson et al., 1999; Narayanan et al., 2000; Lamel et al., 2002). Other work has investigated adding both speech and gesture input (using computer vision) in an interactive kiosk (Wahlster, 2003; Cassell et al., 2002).</Paragraph>
    <Paragraph position="3"> We describe MATCHKiosk, (Multimodal Access To City Help Kiosk) an interactive public information kiosk with a multimodal interface which provides users with the flexibility to provide input using speech, handwriting, touch, or composite multimodal commands combining multiple different modes. The system responds to the user by generating multimodal presentations which combine spoken output, a life-like graphical talking head, and dynamic graphical displays. MATCHKiosk provides an interactive city guide for New York and Washington D.C., including information about restaurants and directions on the subway or metro.</Paragraph>
    <Paragraph position="4"> It develops on our previous work on a multimodal city guide on a mobile tablet (MATCH) (Johnston et al., 2001; Johnston et al., 2002b; Johnston et al., 2002a). The system has been deployed for testing and data collection in an AT&amp;T facility in Washington, D.C. where it provides visitors with information about places to eat, points of interest, and getting around on the DC Metro.</Paragraph>
  </Section>
class="xml-element"></Paper>