<?xml version="1.0" standalone="yes"?>
<Paper uid="W03-1813">
  <Title>Licensing Complex Prepositions via Lexical Constraints</Title>
  <Section position="2" start_page="0" end_page="0" type="intro">
    <SectionTitle>
1 Introduction
</SectionTitle>
    <Paragraph position="0"> Among numerous types of expressions that seem unpredictable regarding standard grammar regularities, sequences consisting of a preposition, a noun, and another preposition (P a0 N a0 Pa1 ) are particularly frequent. Interestingly, this class of expressions, usually termed in linguistic literature as &amp;quot;complex prepositions&amp;quot; (CPs), &amp;quot;phrasal prepositions&amp;quot;, &amp;quot;quasi-prepositions&amp;quot; or &amp;quot;preposition-like word formations&amp;quot; occurs in many different languages, thereby showing nearly uniform properties (cf.</Paragraph>
    <Paragraph position="1"> examples below quoted from (Lindqvist, 1994), (Quirk and Mulholland, 1964), (Grochowski et al., 1984) and (Benes, 1974)).</Paragraph>
    <Paragraph position="2"> French: en face de, en depit de, au milieu de Spanish: al lado de, en casa de Swedish: i borjan av, med hjalp av, i stallet for English: in view of, in spite of, by dint of Polish: bez wzgl,edu na, w zwi ,azku z, z uwagi na German: an Hand von,, mit Hilfe von, in Bezug auf Traditionally, CPs are assumed to be complex lexical categories evincing prepositional character. As well as in the case of other multiword expressions (MWEs), the question arises of how they should be analyzed to make them suitable for machine processing.1 null In this paper, we will propose an HPSG-based linguistically motivated, formal treatment of CPs, applicable for computational platforms intended for developing typed feature structure grammars.</Paragraph>
    <Paragraph position="3"> The starting point of our investigations is the summary of empirical facts about CPs. Although, we have indicated above that CPs can be considered a cross-linguistic phenomenon, we will focus in this paper exclusively on German data, because they provide very explicit and convincing linguistic evidence which motivates and supports our approach. However, we assert that the analysis proposed here for  such as listing &amp;quot;words with spaces&amp;quot;, hierarchically organized lexicons, restricted combinatoric rules, lexical selection, &amp;quot;idiomatic constructions&amp;quot; and simple statistical affinity using HPSG grammar framework see (Sag et al., 2002). On problems of using corpus-based, statistical methods for identifying P a2 N a2 Pa3 word combinations in Dutch referred to in the paper as collocational prepositional phrases (CPPs) see (Bouma and Villada, 2002).</Paragraph>
    <Paragraph position="4"> German can also be applied to other languages such as Polish or English.</Paragraph>
  </Section>
class="xml-element"></Paper>