<?xml version="1.0" standalone="yes"?>
<Paper uid="H90-1046">
  <Title>From Task Manager Application Voice Interface Communication Parser Mapper</Title>
  <Section position="2" start_page="227" end_page="228" type="abstr">
    <SectionTitle>
* The CM-SLS incorporates a fast speaker-independent
</SectionTitle>
    <Paragraph position="0"> continuous-speech multiple knowledge-base recognition system. General English models are used to speed up task development, avoiding task-dependent training.</Paragraph>
    <Paragraph position="1">  names in their vocabulary.</Paragraph>
    <Paragraph position="2"> * The CM-SLS offers a high level user interface on a NeXT Machine for efficient end-user access to the applications. We provide an application fi'amework that provides coherence across applications, allows conciseness inside an application, offers an appropriate feedback, and presents a natural structuring of activities allowing a fast and effective access to applications for both casual and expert users.</Paragraph>
    <Paragraph position="3"> * The CM-SLS uses external tools to quickly build new applications. Tools include a case frame grammar compiler, a case flame grammar parser, and a semi-automatic speech knowledge base (used by the RE) generator.</Paragraph>
    <Paragraph position="4"> The Office Manager (OM) To demonstrate our approach to speech interface design, we have implemented the Office Manager system, a system which is meant to provide the user with voice access to a number of common computer-based office applications. The Office Management domain has several interesting properties that make it an ideal instrument for exploring issues in spoken language system design. The critical attributes of this task domain are the following: * It provides a range of interaction requirements, from tight-loop (e.g., calculation) to open-ended (e.g., database retrieval).</Paragraph>
    <Paragraph position="5"> * It focuses on a realistic ta.~k domain that supports meaningful problem-solving activity (e.g., scheduling, information search).</Paragraph>
    <Paragraph position="6"> * It's a domain in which it would be reasonable to expect daily interaction through spoken language. Since the tasks it encompasses are performed regularly, it creates the opportunity to study spoken language interaction on an ongoing basis, under natural conditions. null The Office Manager at present includes the applications listed in Table 1. In addition to the applications themselves, the OM understands a 36 word vocabulary, which is used to execute a variety of control functions, such as creating tasks, switching between them, invoking help, etc. The current (June 1990) implementation of the system in- null eludes a database of addresses for the 111 official participants in the October 1989 Darpa Speech and Natural Language Woxkshop. This database is used by the Voice Mail and PID (Personal Information Directory) components of OM. Our plan is to make available additional databases to users in our environment (for example, a list of department members) and to pursue the development of tools for user-customizable databases. Customization is of two types: the addition or modification of entries in existing databases, and the creation of new, arbitrary databases by the user. Both forms of customization introduce interesting problems for spoken-language systems: the modification of existing recognition knowledge bases (as might be occasioned by the introduction of a new person name), and the creation (by a presumably naive user) of an access language for a new database.</Paragraph>
    <Paragraph position="7"> Conclusion This paper has described a number of innovations in the design of spoken language interfaces. We have advanced a particnlar functional decomposition for the interface and have argued that it identifies key areas in which advances are needed. We have proposed what we believe to be a meaningful metric for system response characteristics. We have also briefly described the Office Management task, which we believe to be particularly suited for the study of spoken language interface issues.</Paragraph>
    <Paragraph position="8"> Our future work includes the development of techniques for structuring recognition and parsing knowledge bases along &amp;quot;object&amp;quot; lines to permit individual applications to inherit language characteristics from theft environment (the OM) and to encourage the modularization and reusability of language components. The goal is to simplify the process of creating languages for particular applications by providing the developer not only with standard interface components but also with standard language components.</Paragraph>
    <Paragraph position="9"> Meaningful study of spoken language interaction requires the use of a system that will be used on a daily basis and whose utility will persist past the initial stages of play and exploration. We believe that the Office Manager is such a system. Systems that do not have this persistence of utility will ultimately have little to tell us about spoken communication with computers.</Paragraph>
  </Section>
class="xml-element"></Paper>