<?xml version="1.0" standalone="yes"?>
<Paper uid="W01-1014">
  <Title>Automatic Augmentation of Translation Dictionary with Database Terminologies in Multilingual Query Interpretation</Title>
  <Section position="3" start_page="2" end_page="3" type="relat">
    <SectionTitle>
2 Related Work
</SectionTitle>
    <Paragraph position="0"> In this paper, we propose to interpret natural language queries in English and Korean with CCGs and argue that word selection problem must be resolved for multilingual query interpretation.</Paragraph>
    <Section position="1" start_page="2" end_page="2" type="sub_section">
      <SectionTitle>
2.1 Combinatory Categorial Grammar
</SectionTitle>
      <Paragraph position="0"> Combinatory Categorial Grammars (CCGs) are combinatory extensions to the categorial grammars (Steedman, 2000). CCGs are among the lexicalized grammars, such as linear indexed grammars and tree adjoining grammars, and are generally known to provide a wide linguistic coverage and a way of processing sentences incrementally. null Table 2 shows the CCG reduction rules proposed for Korean (Park and Cho, 2000). (Steedman, 2000) suggested the reduction rules for English which include backward crossed composition and backward substitution. We adopt this rule set for the processing of the queries in English. null  Example 2 shows a syntactic derivation for an example query using CCG. Transitive verbs like 'wears' are assigned the category (sD2np)/np, which receives a phrase of category np on its right (the second np and the directionality is indicated by the slash /, that is, to the right) and then receives another np on its left (the first np and the directionality is indicated by the backslash D2, that is, to the left), to give rise to the phrase of category s. Such a computation is done by simple function application. Example 3 shows the CCG derivation for a query with coordination.</Paragraph>
      <Paragraph position="1">  In addition to function application utilized in examples 2 and 3, CCGs use rules for a limited set of combinators such as B (function composition), T (type raising), and S (function substitution) to model natural language. The reader is referred to (Steedman, 2000) for further details.</Paragraph>
    </Section>
    <Section position="2" start_page="2" end_page="3" type="sub_section">
      <SectionTitle>
2.2 Multilingual Database Interfaces
</SectionTitle>
      <Paragraph position="0"> There have been many proposals for NLDBs since the 1960's  . In this section, we review some of the more recent ones. (Androutsopoulos et al., 1998; Nelken and Francez, 2000) focus on queries in English with temporal expressions, with a specialized semantic representation language that can handle temporality. Examples are shown below.</Paragraph>
      <Paragraph position="1"> (4) (a) Did any flight circle while runway 2 was open? (b) Which companies serviced BA737 in 1990? (c) During which years did Mary work in marketing? The system in (Klein et al., 1998) interprets noun phrase queries such as 5 in German: (5) Ersatzzeiten wegen Kindererziehung  (Exemption times because of child raising) While the system can analyze noun phrases with various adverbial phrases, it is not reported to handle more complex noun phrase queries such as those with subordinate or coordinate constructions. null None of these work deals with multilingual issues. Nor is there much related work in the field of NLDBs. (Thompson and Mooney, 1999) presents a system that automatically constructs the lexicon for NLDBs, in various languages such as English, Spanish, Japanese, and Turkish, which represents the lexical entries with a pair of the phrases and the corresponding semantic representation in first-order logic. Since the semantic representation for lexical items is determined using the frequency of the general terms of the semantic representation in the corpus made up of the query sentences annotated by their logical representation, the system makes it difficult to incorporate various linguistic considerations on natural language.</Paragraph>
      <Paragraph position="2">  The reader is referred to (Androutsopoulos et al., 1995) for a survey.</Paragraph>
    </Section>
  </Section>
class="xml-element"></Paper>