<?xml version="1.0" standalone="yes"?>
<Paper uid="C88-2151">
  <Title>Using Constraints in a Constructive Version of GPSG</Title>
  <Section position="3" start_page="0" end_page="738" type="intro">
    <SectionTitle>
1 LP = Linear Precedence; FCR = Feature Co-oecmenee Restriction
KIT-FAST (FAST = Functor-Argument Structure for Translation; KIT =
</SectionTitle>
    <Paragraph position="0"> Kilnstliche Intelligenz Lind Textvexsteben = Artificial Intelligence and Text Understanding) as well as ils predecessor KIT-NASEV (NASEV = Neue Analyse-und Syntheseveffahrm zur masehinellea 0bersetzang = New Metheds of Analysis and Synthesis fc~ Machine Translation) constitute the Berlin component of the complementary ~h project of EUROTRA-D, which reeeive~ gnmts by the Federal Minister for Research and Technology under eontra~t 1013211.</Paragraph>
    <Paragraph position="1">  schild/Busemann88\], \[Keller87\], \[Busemann87\] and \[Weisweber87\]). The reader's familarity with the fundamental knowledge of GPSG as presented in \[GKPS85\] will be assumed.</Paragraph>
    <Section position="1" start_page="738" end_page="738" type="sub_section">
      <SectionTitle>
1.1 The Grammar Format
</SectionTitle>
      <Paragraph position="0"> The ID/LP format of the grammar allows the explicit formulation of generalizations about the partial order of the daughters of the ID rules via LP statements. ID rules are tuples of the form ((mother) ---) (daughters)), for example (S ~ {NP, VP}), where (daughters) is a multi-set of categories which can be dominated by the category (mother). Lexical rules are of the form ((mother) ---C/ (wordform)).</Paragraph>
      <Paragraph position="1"> LP statements are of the form '(categoryt) &lt; (categorY2)', for example the LP statement NP &lt; VP requires that the category NP must precede the category VP in a sequence of daughters of any local tree licensed by an ID rule of the grammar.</Paragraph>
      <Paragraph position="2"> For every grammar a set F = {f~, f2 ..... fn} of syntactic features exists with I FI = n. The number 'n' of features can vary from grammar to grammar, but for a particular grammar it is constant.</Paragraph>
      <Paragraph position="3"> Each feature fi has its domain D(fi). A complex category C is an n-tuple with C ~ D(f 1) x D(f2) x ... x D(fn) and the position T of the tuple represents the value of the feature fi.</Paragraph>
      <Paragraph position="4"> Example : F = {N, Vr BAR, PER, PLU, GEN} position i feature fi domain D(f l)</Paragraph>
    </Section>
  </Section>
class="xml-element"></Paper>