<?xml version="1.0" standalone="yes"?>
<Paper uid="H92-1008">
  <Title>Spontaneous Speech Collection for the ATIS Domain with an Aural User Feedback Paradigm</Title>
  <Section position="3" start_page="0" end_page="43" type="intro">
    <SectionTitle>
2. DATA COLLECTION
</SectionTitle>
    <Paragraph position="0"> As at most of the other sites collecting ATIS data, data were collected from subjects' interactions with a partially simulated spoken language system. As in the MIT data collection setup, a human experimenter is substituted for the speech recognition component of the system to provide a transcription of the subject's speech for the natural language (NL) component. In this section, we discuss the system's development, describe the system hardware, and describe the collection procedure.</Paragraph>
    <Section position="1" start_page="0" end_page="43" type="sub_section">
      <SectionTitle>
2.1. System Development
</SectionTitle>
      <Paragraph position="0"> The AT&amp;T data collection system was designed to closely simulate a real, telephone-based, human-machine interaction. Building on the framework of the MIT ATIS collection system, we directed our development effort towards controlling the presentation of information retrieved from the database, providing feedback to the user on the state of the system and the discourse, and exploring areas where system initiative would help users achieve their goals efficiently. We also modified the system control loop to transfer recording control away from the subject.</Paragraph>
      <Paragraph position="1"> Information Presentation. The collection systems at all other sites make use of a visual display to present information retrieved from the database in a tabular format. Because of our choice of an audio interaction paradigm, the AT&amp;T system does not present information in a tabular format, but instead translates the retrieved information into sentences. In some cases, information is converted into sentences using an entry-tophrase, template-based approach (Figure 1). When more information is retrieved from the database than can be reasonably presented by the template-based approach, summarization functions are used to present some sub-set of the information (Figure 2). In other cases, specialized functions are used to avoid excessive repetition (Figure 3) or to select information based on the discourse history.</Paragraph>
      <Paragraph position="2">  TO_AIRPORT ATL to Atlanta DEPARTURE~IME 1520 departs at three tgenty P M ARRIVAL~INE 1804 and arrives at six oh four P M STOPS 0 gith no stops ATIS: delta flight nine seventy five from Boston to Atlanta departs at three tgenty P N and arrives at six oh four P M glth no stops.</Paragraph>
      <Paragraph position="3">  I MANT TO GO FROM BOSTON TO ATLANTA ON MONDAY AIRLINE~ODE ... DEPARTURE~IME ...</Paragraph>
      <Paragraph position="4"> DL ... 630 ...</Paragraph>
      <Paragraph position="5"> EA ... 700 ...</Paragraph>
      <Paragraph position="6"> DL ... 815 ...</Paragraph>
      <Paragraph position="7">  Summary: There are flights departing between six thirty A M and eight tgenty four P M.</Paragraph>
      <Paragraph position="8"> ATIS: There are seventeen flights from boston to atlanta on Monday August nineteenth.</Paragraph>
      <Paragraph position="9"> There are flights departing betgeen six thirty A N and eight tgenty four P M.</Paragraph>
      <Paragraph position="10"> What time gould you like to go7</Paragraph>
    </Section>
  </Section>
class="xml-element"></Paper>