<?xml version="1.0" standalone="yes"?>
<Paper uid="W90-0123">
  <Title>Relational-Grammar-Based Generation in the JETS Japanese-English Machine Translation System</Title>
  <Section position="6" start_page="180" end_page="180" type="concl">
    <SectionTitle>
6- Concluding Remarks
</SectionTitle>
    <Paragraph position="0"> We have developed an application-and-NL-independent generator shell, GENSHELL, including a flexible dictionary system and a high-level rule-writing system, GEAR, to facilitate the development of category-driven RG generators. GENSHELL/GEAR provides a powerful computational framework for the development of RG-based natural-language-processing components.</Paragraph>
    <Paragraph position="1"> We have also implemented GENIE, a robust English generator, within GENSHELL/GEAR. Besides the novel use of RG and category-driven processing, GENIE is notable for its two-stage plan-and-execute design.</Paragraph>
    <Paragraph position="2"> JETS and GENIE are currently being tested on sentences from Asahi newspaper editorials on economic matters, a challenging task since editorial sentences can be very long, with essentially unrestricted vocabulary. Nevertheless, we have found the initial tests of the generator encouraging, supporting our view that besides its intrinsic theoretical interest, RG has practical value in natural language processing.</Paragraph>
  </Section>
class="xml-element"></Paper>