<?xml version="1.0" standalone="yes"?>
<Paper uid="C82-2031">
  <Title>DEEP CASES FOR ADVERBIAL AND PREPOSITIOI~AL PHRASES</Title>
  <Section position="1" start_page="0" end_page="0" type="abstr">
    <SectionTitle>
DEEP CASES FOR ADVERBIAL AND PREPOSITIOI~AL PHRASES
Lieven Jaspaert
</SectionTitle>
    <Paragraph position="0"> Section of Applied Linguistics, Katholieke Universttett Leuven Blinds Inkomststraat 21, 3000 Leuven, Belgium This paper attempts to demonstrate the usefulness for Machine Translation of a case-oriented description of the syntactic and the semantic behavior of adverbial and prepositional constructions.</Paragraph>
    <Paragraph position="1"> Ee aim to modify and refine the existing descriptiohs of adverbs and prepositional phrases in order to put them to efficient use in Machine Translation. We hope to point out, furthermore, the desirability of incorporating deep case information into the &amp;quot;interface structure&amp;quot; of a Machine Translation system.</Paragraph>
    <Paragraph position="2"> I. Task-svecific requirements on parsing Parsing sentences for translation purposes is bound to differ i~ significant ways from parsing natural language for other tasks. In this paragraph, we will formulate a n,,mbar of translation-speclfio requirements on natural language analysis which, moreover, will be of particular interest in the discussion of the issues in the following paragraphs.</Paragraph>
    <Paragraph position="3"> 2. Surface cases, deep cases and conceptual caqes A much debated issue in AI is the desirable depth of the case system employed In natural language understanding programs. From the preceding paragraph, however, it will have become apparent that, in Machine Translation, one could better - 137 make use of various case systems of increasing depth. In this paragraph, we will be concerned with the definition of the three types of oases we consider to be adequate for translational purposes: (i) surface cases reflect the dependency relation~ which exist between constituents and their governors; (ii) deep cases provide information about the valency boundnese of the participants in the predication; (ili) conceptusl cases are language-independent values which give information about the states of affairs people talk about.</Paragraph>
    <Paragraph position="4"> 3. Deep ca~eq for adverbial and prepositional phrases It is common practice in valency grammar to assign the properties + valency-bound or - valency-bound to constituents. We hope to show that this dichotomy is inadequate for dealing with natural language in general, and with adverbial end prepositional phrases in particular. The discussion will aim at introducing a third property to the set of deep cases ccmnonly used: the &amp;quot;trenspropositional case deg.</Paragraph>
    <Paragraph position="5"> The main part of this Section will be devoted to the definition of the transpropositional case. We will point out the descriptive and computational advantages of the three-case system proposed, on the basis of examples. This will lead to a classification of adverbial and prepositional constructions which differs in a considerable way from the one traditionally used in linguistics. F~ther attention will be expended to constructions other than adverbial and prepositional ones, which should as well be labeled trsnspropoeitional.</Paragraph>
  </Section>
class="xml-element"></Paper>