<?xml version="1.0" standalone="yes"?>
<Paper uid="A88-1004">
  <Title>TIHE PRESENT :SYHPTOH (*PAIN :LOCATION (*BOOY-PART :NAHE *THROAT</Title>
  <Section position="11" start_page="29" end_page="30" type="concl">
    <SectionTitle>
6.2 Related work
</SectionTitle>
    <Paragraph position="0"> There is more recent work in &amp;quot;multilingual generation&amp;quot; from data: RAREAS, a system synthesizing weather forecasts from data provided by meteorologists \[Kittredge et al. 86\] is currently being equipped with French as second target language.</Paragraph>
    <Paragraph position="1"> Kukich's ANA, a system generating English stock market reports from Dow Jones data \[Kukich 83\], has a second tongue as well: The generation of French bulletins has been possible by replacing ANA's &amp;quot;linguistic module&amp;quot; with a French version - called FRANA \[Contant 86\] while leaving ANA's other modules untouched (i.e. Fact Generator, Message Generator, Discourse Organizer).</Paragraph>
    <Section position="1" start_page="29" end_page="30" type="sub_section">
      <SectionTitle>
6.3 Why multilingual generation?
6.3.1 Aspects of application:
</SectionTitle>
      <Paragraph position="0"> Generation of natural language texts in different languages (and probably different styles) from the same knowledge base might be an  interesting alternative to human or machine translation of these texts.</Paragraph>
      <Paragraph position="1"> Re-Generation (of e.g. software manuals or maintenance handbooks) in different languages might be much more economic than manually &amp;quot;updating&amp;quot; those texts when the underlying knowledge base changes.</Paragraph>
      <Paragraph position="2">  Multilingual generation enforces the separation of generator knowledge into language dependent data and language independent machinery. null In order to keep a generator easily portable to other languages the implementor will have to allow as much declarativity as possible.</Paragraph>
    </Section>
    <Section position="2" start_page="30" end_page="30" type="sub_section">
      <SectionTitle>
6.3.3 Aspects of linguistic theory:
</SectionTitle>
      <Paragraph position="0"> Work in multilingual generation from semantic representations may be seen as an exercise in contrastive linguistics: A central issue for any generator starting from semantic structures is the choice of an appropriate syntactic structure for the expression of a given meaning structure.</Paragraph>
      <Paragraph position="1"> What are the differences and correspondencies between the different target languages with respect to this choice? (E. g. Focus may be expressed by constituent order in German, in English you may have to choose a passive.) Similarly.</Paragraph>
      <Paragraph position="2"> What are the differences between the target languages with respect to the semantic features that are obligatory in order to be able to produce surface text? (E.g.</Paragraph>
      <Paragraph position="3"> In the Japanese/German MT application we were confronted with the fact that the semantic structures derived from Japanese did not contain semantic information about in/definiteness or multiplicity.) ACKNOWLEDGEMENTS: The SEMSYN project is funded by the West German Ministry for Research and Technology (BMFT). The project is currently cooperating with partners from Japan (University of Kyoto; NTT Basic Research Laboratories) and USA (International Center for MT at CMU).</Paragraph>
      <Paragraph position="4"> We have to thank all partners for their support. The SEMSYN system is the joint effort of a variety of people. Special thanks to M. Emele (Stuttgart) for his work on the front end generator, to W. Kehl (Stuttgart) for his implementation of GEOTEX and the editor for semantic nets and to O. Rambow (Ithaca, N.Y.) for our joint experiment to teach English to the system.</Paragraph>
      <Paragraph position="5"> A note on implementation and demonstration The SEMSYN generator and the applications as described in this paper are fully implemented and run in ZetaLISP and FLAVORS on SYMBOLICS lisp machines. We would like to demonstrate the system at the conference.</Paragraph>
    </Section>
  </Section>
class="xml-element"></Paper>