<?xml version="1.0" standalone="yes"?>
<Paper uid="P83-1018">
  <Title>SYN'I'ACI IC CONSTI~,,\INTS AND F~FI:ICIFNI' I~AI(SAI~,II.I'I'Y</Title>
  <Section position="1" start_page="0" end_page="0" type="metho">
    <SectionTitle>
SYN'I'ACI IC CONSTI~,,\INTS AND F~FI:ICIFNI' I~AI(SAI~,II.I'I'Y
</SectionTitle>
    <Paragraph position="0"/>
  </Section>
  <Section position="2" start_page="0" end_page="0" type="metho">
    <SectionTitle>
ABSTRACT
</SectionTitle>
    <Paragraph position="0"> A central goal of linguistic theory is to explain why natural languages are the way they are. It has often been supposed that com0utational considerations ought to play a role in this characterization, but rigorous arguments along these lines have been difficult to come by. In this paper we show how a key &amp;quot;axiom&amp;quot; of certain theories of grammar, Subjacency, can be explained by appealing to general restrictions on on-line parsing plus natural constraints on the rule-writing vocabulary of grammars. The explanation avoids the problems with Marcus' \[1980\] attempt to account for the same constraint. The argument is robust with respect to machine implementauon, and thus avoids the problems that often arise wilen making detailed claims about parsing efficiency. It has the added virtue of unifying in the functional domain of parsing certain grammatically disparate phenomena, as well as making a strong claim about the way in which the grammar is actually embedded into an on-line sentence processor.</Paragraph>
  </Section>
  <Section position="3" start_page="0" end_page="119" type="metho">
    <SectionTitle>
I INTRODUCTION
</SectionTitle>
    <Paragraph position="0"> In its short history, computational linguistics has bccn driven by two distinct but interrelated goals. On the one hand, it has aimed at computational explanations of distinctively human linguistic behavior -- that is, accounts of why natural languages are the way they are viewed from the perspective of computation. On the other hand, it has accumulated a stock of engineenng methods for building machines to deal with natural (and artificial) languages. Sometimes a single body of research has combined both goals. This was true of the work of Marcus \[1980\]. for example. But all too often the goals have remained opposed -- even to the extent that current transformational theory has been disparaged as hopelessly &amp;quot;intractable&amp;quot; and no help at all in constructing working parsers.</Paragraph>
    <Paragraph position="1"> This paper shows that modern transformational grammar (the &amp;quot;Government-Binding&amp;quot; or &amp;quot;GB&amp;quot; theory as described in Chomsky \[1981\]) can contribute to both aims of computational linguistics. We show that by combining simple assumptions about efficient parsability along with some assumpti(ms about just how grammatical theory is to be &amp;quot;embedded&amp;quot; in a model of language processing, one can actually explain some key constraints of natural languages, such as Suhjacency.</Paragraph>
    <Paragraph position="2"> (The a)gumcnt is differmlt frt)m that used in Marcus 119801.) In fact, almost the entire pattern of cunstraints taken as &amp;quot;axioms&amp;quot; by the GB thct)ry can be accutmtcd tbr. Second, contrary to what has sometimes been supposed, by exph)iting these constraints wc can ~how that a Gll-based theory is particularly compatil)le v~idl efficient parsing designs, in particdlar, with extended I I~,(k,t) parsers (uf the sort described by Marcus \[1980 D. Wc can extcnd thc I,R(k.t) design to accommodate such phenomena as antecedent-PRO and pronominal binding. Jightward movement, gappiug, aml VP tlcletion.</Paragraph>
    <Paragraph position="3"> A, Functional Explanations o__f I,ocality Principles Let us consider how to explain locality constraints in natural languages. First of all, what exactly do we mean by a &amp;quot;locality constraint&amp;quot;? &amp;quot;\]'he paradigm case is that of Subjacency: the distance between a displaced constituent and its &amp;quot;underlying&amp;quot; canonical argument position cannot be too large, where the distance is gauged (in English) in terms of the numher of the number of S(entence) or NP phrase boundaries. For example, in sentence (la) below, John (the so-called &amp;quot;antecedent&amp;quot;) is just one S-boundary away from its presumably &amp;quot;underlying&amp;quot; argument position (denoted &amp;quot;x&amp;quot;, the &amp;quot;trace&amp;quot;)) as the Subject of the embedded clause, and the sentence is fine: (la) John seems \[S x to like ice cream\].</Paragraph>
    <Paragraph position="4"> However, all we have to do ts to make the link between John and x extend over two S's, and the sentence is ill-formed: (lb) John seems \[S it is certain \[S x to like ice cream This restriction entails a &amp;quot;successive cyclic&amp;quot; analysis of transformational rules (see Chomsky \[1973\]). In order to derive a sentence like (lc) below without violating the Subjacency condition, we must move the NP from its canonical argument position through the empty Subject position in the next higher S and then to its surface slot: (lc) John seems tel to be certain x to get the ice cream.</Paragraph>
    <Paragraph position="5"> Since the intermediate subject position is filled in (lb) there is no licit derivation for this sentence.</Paragraph>
    <Paragraph position="6"> More precisely, we can state the Subjacency constraint as follows: No rule of grammar can involve X and Y in a configuration like the following, \[ ...x...\[,, ...\[/r..Y...\]...l ...X...\] where a and # are bounding nodes (in l.'nglish, S or NP phrases). &amp;quot; Why should natural languages hc dcsigned Lhis way and not some other way? Why, that is, should a constraint like Subjaccncy exist at all? Our general result is that under a certain set of assumptions about grammars and their relationship to human sentence processing one can actually expect the following pattern of syntactic igcality constraints: (l) The antecedent-trace relationship must obey Subjaccncy, but other &amp;quot;binding&amp;quot; realtionships (e.g., NP--Pro) need not obey Subjaccncy.</Paragraph>
    <Paragraph position="7">  (2) Gapping constructitms must be subject to a bounding condition resembling Subjacency. but VP deletion nced not be.</Paragraph>
    <Paragraph position="8"> (3) Rightward movemcnt must be stricdy  bounded.</Paragraph>
    <Paragraph position="9"> To the extent that this predicted pattern of constraints is actually observed -- as it is in English and other languages -- we obtain a genuine functional explanation of these constraints and support for the assumptions themselves. The argument is different from Man:us' because it accounts for syntactic locality constraints (like Subjaceney) ,as the joint effect of a particular theory of grammar, a theory of how that grammar is used in parsing, a criterion for efficient parsability. and a theory of of how the parser is builL In contrast, Marcus attempted to argue that Subjaceney could be derived from just the (independently justified) operating principles of a particular kind of parser.</Paragraph>
    <Paragraph position="10"> B. Assumptions.</Paragraph>
    <Paragraph position="11"> The assumptions we make are the following: (1) The grammar includes a level of annotated surface structure indicating how constituents have been displaced from their canonical predicate argument positions.</Paragraph>
    <Paragraph position="12"> Further, sentence analysis is divided into two stages, along the lines indicated by tile theory of Government and Binding: the first stage is a purely syntactic analysis that rebuilds annotated surface structure; the second stage carries out the interpretation of variables, binds them to operators, all making use of the &amp;quot;referential indices&amp;quot; of NPs.</Paragraph>
    <Paragraph position="13"> (2) To be &amp;quot;visible&amp;quot; at a stage of analysis a linguistic representation must be written in the vocabulary of that level. For example, to be affected by syntactic operations, a representation must be expressed in a syntactic vocabulary (in the usual sense); to be interpreted by operations at the second stage, the NPs in a representation must possess referential indices. (This assumption is not needed to derive the Subjaccncy constraint, but may be used to account for another &amp;quot;axiom&amp;quot; of current grammatical theory, the so-called &amp;quot;constituent command&amp;quot; constraint on antecedcnLs and the variables that they hind.) This &amp;quot;visibility&amp;quot; assumption is a rather natural one.</Paragraph>
    <Paragraph position="14"> (3) The rule-writing vocabulary of the grammar cannot make use of arithmetic predicates such as &amp;quot;one&amp;quot;, &amp;quot;two&amp;quot; or &amp;quot;three&amp;quot;. but only such predicates as &amp;quot;adjacent&amp;quot;.</Paragraph>
    <Paragraph position="15"> Further, quzmtificational statements are not allowed m rt.les. These two assumptions are also rather standard. It has often been noted that grammars &amp;quot;do not count&amp;quot; -- that grammatical predicates are structurally based. There is no rule of grammar that takes the just the fourth constituent of a sentence and moves it, for example. In contrast, many different kinds of rules of grammar make reference to adjacent constituents. (This is a feature found in morphological, phonological, and syntactic rules.) (4) Parsing is no....! done via a method that carries along (a representation) of all possible derivations in parallel. In particular, an Earley-type algorithm is ruled out. To the extent that multiple options about derivations are not pursued, the parse is &amp;quot;deterministic.&amp;quot; (5) The left-context of the parse (as defined in Aho and Ullman \[19721) is literally represented, rather than generatively represented (as, e.g., a regular set). In particular, just the symbols used by the grammar (S, NP. VP...) are part of the left-context vocabulary, and not &amp;quot;complex&amp;quot; symbols serving as proxies for the set of lefl.-context strings. 1 In effect, we make the (quite strong) assumption that the sentence processor adopts a direct, transparent embedding of the grammar.</Paragraph>
    <Paragraph position="16"> Other theories or parsing methods do not meet these constraints and fail to explain the existence of locality constraints with respect to thts particular set of assumpuons. 2 For example, as we show, there is no reason to expect a constraint like Subjacency in the Generalized Phrase Structure Grammars/GPSGsl of G,zdar 119811, because there is no inherent barrier to eastly processing a sentence where an antecedent and a trace are !.mboundedly far t'rt~m each other. Similarly if a parsing method like Earlcy's algorithm were actually used by people, than Sub\]acency remains a my:;tcry on the functional grounds of efficient parsability. (It could still be explained on other functional grounds, e.g., that oflearnability.)</Paragraph>
  </Section>
class="xml-element"></Paper>