<?xml version="1.0" standalone="yes"?>
<Paper uid="C92-2100">
  <Title>A Three-level Revision Model for Improving Japanese Bad-styled Expressions</Title>
  <Section position="3" start_page="0" end_page="0" type="intro">
    <SectionTitle>
2 Computational Aids to Revision
</SectionTitle>
    <Paragraph position="0"/>
    <Section position="1" start_page="0" end_page="0" type="sub_section">
      <SectionTitle>
2.1 Targets of the Computational Aids
</SectionTitle>
      <Paragraph position="0"> Misldmal6\] has summarized the key conditions for efficient technical communication via texts as follows:  (1) The reader must be able to easily understand the text (Easy-understanding).</Paragraph>
      <Paragraph position="1"> (2) The readers must be able to correctly understaod the text (Correct-understanding).</Paragraph>
      <Paragraph position="2"> (3) The contents of tile text nmst meet tile ieader's  purposes.</Paragraph>
      <Paragraph position="3"> Front this viewpomt+ the task of revision is text rewriting or regenerating to make the original text satisfy these conditions. The last condition is too hard to support computatiooatly; however, the first two conditions are promising because we can apply natural language processing technologies. Therefore, we concentrate on tile first two conditions in designing the colnputer~assisted revision systen/.</Paragraph>
    </Section>
    <Section position="2" start_page="0" end_page="0" type="sub_section">
      <SectionTitle>
2.2 Revising as Regeneration or Rewriting
</SectionTitle>
      <Paragraph position="0"> Is there just one computational revision model which is suitable for use?. Two models illustrated in Fig.l are the two extrenles and will give the basis for constructing u practical model: one is tile Regeneration based model (a) and the other is the Rewriting-based model (b).</Paragraph>
      <Paragraph position="2"> The regeneration-based model is a strong model; namely, if all of its components are perfectly constructed, the output text will be understandable and contain no badly-styled expressions. IIowever, if some component is incomplete, there is a possibility of some inpet text being fatally flawed. Therefore, for this model to be practical, as a minimum, the following two major problems olust be overcome:</Paragraph>
    </Section>
  </Section>
class="xml-element"></Paper>