<?xml version="1.0" standalone="yes"?>
<Paper uid="C82-1034">
  <Title>MULTI-LEVEL TRANSLATION AIDS IN A DISTRIBUTED SYSTEM</Title>
  <Section position="2" start_page="0" end_page="215" type="intro">
    <SectionTitle>
OVERVIEW
</SectionTitle>
    <Paragraph position="0"> At COLING80, we reported on an Interactive Translation System ealled ITS. We will consider three problems in the first version of ITS: (1) human factors, (2) the 'Wall or nothing&amp;quot; syndrome, and (3) traditional centralized processing.</Paragraph>
    <Paragraph position="1"> The first problem (human factors) is the problem of keeping human translators and revisors happy. Humans naturally want to feel that they are doing useful, interesting work and that they are using the machine instead of it using them. However, the first version of ITS forced them to answer many uninteresting questions and to revise many sentences they thought should be retranslated.</Paragraph>
    <Paragraph position="2"> The &amp;quot;el! or nothing&amp;quot; syndrome is a name for the attitude that the machine must translate every sentence or it is not worth using a machine at all The problem is that a system based on this approach is likely to be hard to adjust into a useful form if it does not attain the desired level of performance.</Paragraph>
    <Paragraph position="3"> The problem with traditional centralized processing is that it does not provide consistent, reliable response time to e~ch user and requires physical proximity or high-speed telecommunications. And a eentralized system may be hard to decentralize after it has been designed.</Paragraph>
    <Paragraph position="4"> The first version of ITS had all three of the above problems. These problems would disappear if we had FAI-IQT (Fully Automatic, High Quality Translation -- Barllillel,1960). In that case a source text would be presented to the computer, which would promptly produce a poIished translation, typeset and ready to be published without revision. That would solve the human problems because no human translators would be involved. The &amp;quot;all or nothing&amp;quot; question would be irrelevant because we would have it all. And centralized processing would not be a problem because there would be no interactive processing. This paper assumes that FAllQT of general text is not on the visible horizon and proposes a design which answers these problems.</Paragraph>
    <Paragraph position="5"> In the new version of ITS, each translator works at a microcomputer instead of a conventional terminal. The microcomputers are part of a distributed network but can function without being on-line. The translator uses the microcomputer as a tool for getting the translation done and is in control of the translation process. There are three levels of aid available to the translator, ranging from simple text processing to</Paragraph>
  </Section>
class="xml-element"></Paper>