<?xml version="1.0" standalone="yes"?>
<Paper uid="J92-2001">
  <Title>Inheritance in Word Grammar</Title>
  <Section position="4" start_page="141" end_page="143" type="intro">
    <SectionTitle>
3. Syntax
</SectionTitle>
    <Paragraph position="0"/>
    <Section position="1" start_page="141" end_page="143" type="sub_section">
      <SectionTitle>
3.1 Word Types
</SectionTitle>
      <Paragraph position="0"> WG syntax is centred on two inheritance hierarchies, one for word types (i.e. word classes and lexical items) and the other for grammatical relations. In Word Grammar  Norman M. Fraser and Richard A. Hudson Inheritance in Word Grammar word adword noun conjunct ion verb adjective adverb proper common pronoun polarity STAND etc preposition Figure 1 The word type hierarchy.</Paragraph>
      <Paragraph position="1"> count determiner modal (as suggested by the name) the category 'word' is basic in every sense. Figure 1 shows the top of the hierarchy of word types assumed in WG for English, and some of the  corresponding WG propositions are given in (27). (27a) count isa common.</Paragraph>
      <Paragraph position="2"> (27b) common isa noun.</Paragraph>
      <Paragraph position="3"> (27c) noun isa word.</Paragraph>
      <Paragraph position="4"> Three points should be noted about this hierarchy. 1. We assume a hierarchical relation among word types, instead of the more usual cross-classification based on features. This links to a general restriction on the use of features in WG, which excludes all features except those that are morphosyntactic--i.e., reflected in morphology and relevant to syntax or semantics.</Paragraph>
      <Paragraph position="5"> 2. Secondly, we assume some nonstandard analyses; in particular, a preposition is a kind of adverb, and a determiner is a kind of pronoun, which in turn is a kind of noun.</Paragraph>
      <Paragraph position="6"> 3. We prefer to keep an open mind on the extent to which our categories are universal, but we are sure that some are parochial (relevant to English only). This hierarchy can be continued downward to include lexical items (such as STAND, shown in the diagram), which may in turn be further subdivided; e.g., we can distinguish transitive and intransitive versions of STAND (with, it should be noted, the same irregular morphology in both cases): (28a) STAND isa verb.</Paragraph>
      <Paragraph position="7"> (28b) STAND/intrans isa STAND.</Paragraph>
      <Paragraph position="8"> (28c) STAND/trans isa STAND.</Paragraph>
      <Paragraph position="9">  Computational Linguistics Volume 18, Number 2 As explained earlier, because lexical items are part of the same hierarchy as general word classes, there is no formal distinction between the lexicon and the rest of the grammar. Furthermore, we use the same isa relation to link word tokens to word types; so if w3 is the name of the word stand in I can't stand cats, it too will fit into the same hierarchy: (29) w3 isa STAND/trans.</Paragraph>
      <Paragraph position="10"> Word tokens can be thought of as a constantly changing fringe on the bottom of the (relatively) permanent hierarchy.</Paragraph>
    </Section>
    <Section position="2" start_page="143" end_page="143" type="sub_section">
      <SectionTitle>
3.2 Grammatical Functions
</SectionTitle>
      <Paragraph position="0"> We now come to the second hierarchy of syntax, the hierarchy of grammatical relations. Unlike most other syntactic theories, WG uses constituent structure only for the purpose of describing coordinate constructions (cf. Hudson 1990: 404ff for details). All other syntactic structure is expressed in terms of dependencies between pairs of words, one of which is the head of the other, its dependent. Higher nodes such as phrases or sentences are not represented explicitly in the grammar. WG is thus a variety of dependency grammar.</Paragraph>
      <Paragraph position="1"> Dependency grammar was first formalized by Tesni6re (1959) and refined by Hays (1964), Gaifman (1965), Robinson (1970) and others. A number of dependency-based theories have emerged from the linguistic underground during the last thirty years. These include the Meaning-Text model (Mel'~uk and Zolkovskij 1970; Mel'~uk 1988), Case Grammar (Anderson 1971; 1977), Daughter Dependency Grammar (Hudson 1976), WG (Hudson 1984; 1990), Functional Generative Description (Sgall, Haji~ov~, and Panevov~ 1986), and Lexicase (Starosta 1988). While none of these theories has attained widespread popularity, some of their central insights have become increasingly influential in the phrase structure grammar mainstream. For example, the trend toward head-driven approaches, the prominence of notions such as 'government,' the explicit use of grammatical relations and case, and the reduced amount of information carried in phrasal categories all reflect the general migration toward dependency. Increased interest in categorial grammars, and especially unification categorial grammars (which are virtually indistinguishable from dependency grammars) provides further evidence of this tendency. 3 The combination of default inheritance with dependency syntax allows an interesting range of generalizations and exceptions. Like other dependency grammars, WG requires a typical word to have one head, though the same word may act as head to more than one other word, its dependents. As in other theories, just one word is allowed to be an exception to this rule; we call this word the 'root' of the sentence. This has (by definition) no head, and is generally a finite verb; e.g. in Mary didn't jump, the polarity verb ('auxiliary verb') didn't is the root, on which both Mary and jump</Paragraph>
    </Section>
  </Section>
class="xml-element"></Paper>