LINGUIST List 11.2540

Sat Nov 25 2000

Review: Steedman: The Syntactic Process

Editor for this issue: Andrew Carnie <carnielinguistlist.org>



What follows is another discussion note contributed to our Book Discussion Forum. We expect these discussions to be informal and interactive; and the author of the book discussed is cordially invited to join in. If you are interested in leading a book discussion, look for books announced on LINGUIST as "available for discussion." (This means that the publisher has sent us a review copy.) Then contact Andrew Carnie at carnielinguistlist.org

Directory

  • Radu Daniliuc, Review: Mark Steedman: The Syntactic Process

    Message 1: Review: Mark Steedman: The Syntactic Process

     Date: Wed, 22 Nov 2000 12:29:55 -0800
    From: Radu Daniliuc <radu.daniliucanu.edu.au>
    Subject: Review: Mark Steedman: The Syntactic Process


    Mark Steedman: The Syntactic Process. A Bradford Book 2000, MIT Press Language, Speech, and Communication Series 330 pages

    Reviewed by Laura and Radu Daniliuc (The Australian National University)

    At the end of a century that experienced a real boom of syntactic theories, MIT Press offers its voracious readers a new book presenting a fresh perspective on the theory of grammar. Its author is Mark Steedman, Professor of Cognitive Science in the Division of Informatics at the University of Edimburgh. He is also the author of "Surface Structure and Interpretation" (MIT Press, 1996), which describes a new approach to the theory of natural language grammar, namely the Combinatory Categorial Grammar. The present book, published in the Language, Speech, and Communication Series, develops and clarifies some of the ideas previously formulated and it is built on the argument that the surface syntax of natural languages maps spoken and written forms directly to a compositional semantic representation including predicate-argument structure, quantification and information structure without forming any intervening structural representation. Steedman's theory of grammar belongs to the wider circle of Categorial Grammar, formed by a group of theories of natural language, in which the main responsibility for defining syntactic form is borne by the lexicon. Categorial Grammar is considered to be one of the oldest and purest examples of lexicalized theories of grammar that also includes head-driven phrase structure grammar, lexical functional grammar, Tree-adjoining grammar, Montague grammar, relational grammar, and certain recent versions of Chomsky 's theory. Since its beginning somewhere in the 30's (Ajdukiewicz, 1935), Categorial Grammar has aimed at describing a language by assigning logical types (syntactic or semantic expressions) to lexical atoms. Types of complex expressions are derived from types of atomic expressions by a logical proof that relies on fundamental logical rules interpreted in terms of type theory. The most important variants of Categorial Grammar were proposed in the eighties, the most representative being Steedman's Combinatory Categorial Grammar, as mainly explained in "The Syntactic Process". Drawing on ideas explored by other categorical grammars, his theory does not cover a narrow direction, but it refers to such areas as formal linguistics, intonational phonology, computational linguistics, and experimental psycholonguistics and is orientated toward Chomsky's goal of formalizing an explanatory theory of linguistic form. Steedman tries to demystify the complexity of syntax by arguing that this complexity is the result of the lexical specification of grammar and of the small number of universal rule-types for combining predicates and arguments of the appropriate type and position by rules of functional application written as X/Y Y => X and Y X\Y => X. In order to associate predicate-argument structures with syntactic categories, the functional application rules must be expanded as X/Y:f Y:a => X:Ya and as Y:a X\Y:f => X:f a (p.37), rules that apply an identical compositional-semantic operation in both syntax and semantics. Steedman considers that all combinatory rules permitted in Combinatory Categorial Grammar obey the Principle of Combinatory by Transparency which states "all syntactic combinatory rules are type-transparent versions of one of a small number of simple semantic operations over functions" (p.37) and their involvement offers a common mechanism for canonical word order, leftward extraction constructions and right-node-raising constructions based on a single lexical entry for the verb. As it can easily be deduced, the theory of Combinatory Categorial Grammar, which tries to show a harmony between syntax, semantics, phonology, and discourse information, is most closely related to Generalized Phrase Structure Grammar (Gazdar 1981), Head Grammar (Pollard 1984) and Tree-Adjoining Grammar (Joshi, Levy and Takahashi 1975). Its most basic assumption has been formulated as the Principle of Lexical Head Government which states that "both bounded and unbounded syntactic dependencies are specified by the lexical syntactic type of their head" (p.32). This lexicalized view on grammar is shaped by the Principle of Head Categorial Uniqueness which minimizes the size of the lexicon involved by arguing that "a single nondisjunctive lexical category for the head of a given construction specifies both the bounded dependencies that arise when its complements are in canonical position and the unbounded dependencies that arise when these complements are displaced under relativization, coordination, and the like" (p.33). The directionality specified in the lexicon may not be contradicted by combinatory rules, which are characterized in terms of three universal principles delimiting the space of possible combinatory rules in all human languages. These principles are: 1. The Principles of Adjacency: Combinatory rules may only apply to finitely many phonologically realized and string-adjacent entities. 2. The Principle of Consistency: All syntactic combinatory rules must be consistent with the directionality of the principal function. 3. The Principle of Inheritance: If the category that results from the application of a combinatory rule is a function category, then the slash defining directionality for a given argument in that category will be the same as the one(s) defining directionality for the correspondent argument(s) in the input function(s). (p.54) These are the general ideas underlying Steedman's theory of grammar as they are presented in his book. As for its structure, "The Syntactic Process" comprises three parts, the first two ("Grammar and Information Structure" and "Coordination and Word Order") entirely centered on competence and the last one ("Computation and Performance") dealing with issues of performance mechanisms and computational matters. One of the main assumptions of the book is that "the Surface Syntax of natural language acts as a completely transparent interface between the spoken form of the language, including prosodic structure and intonational phrasing, and a compositional semantics" . (p.1) The syntactic and semantic components are related in Steedman's theory by the Principle of Categorial Type Transparency according to which "for a given language, the semantic type of the interpretation together with a number of language-specific directional parameter settings uniquely determines the syntactic category of a category" (p.36). The book, considered by its author an attempt to reunite in "a single framework and a uniform notation" the results of the project that he and his colleagues worked on, begins by stating some uncontroversial assumptions in the form of rule-to-rule condition and the competence hypothesis, deducing some of the even more widely accepted Constituent Condition on rules of competence grammar. The Introduction endorses the methodological priority of investigating complex syntax over performance mechanisms. Part I, which advances an alternative combinatory view of competence grammar, begins with a chapter on rules, constituents, and fragments and offers a rethinking of the nature of Surface Structure from coordination, parenthicalization, and intonation. The traditional notion of Surface Structure is entirely replaced by a freer notion ("the only necessary derivational one") of "surface constituency" corresponding to Information Structure. Syntactic structure is merely "the characterization of the process of constructing a logical form, rather than a representational level of structure that actually needs to be built". (p.xi) The next chapter presents the intuitive basis of Combinatory Categorial Grammars with simple examples motivating the individual rule types. Furthermore, Steedman explains the constraints on Natural Grammar, i.e. constraints on bounded and unbounded constructions. Chapter 5 focuses on structure and intonation, revising and extending the author's former views (1991) on the matter. Prosodic information is integrated with the standard grammatical categories to more directly capture Intonation Structure, together with its interpretation as Information Structure. The reader should be reminded that in Combinatory Categorial Grammar Intonation Structure and Surface Structure are subsumed under the notion of Information Structure and that Intonation Structure and discourse Information Structure are integrated into the grammar itself. Part II can be perceived as a more technical approach that draws on two related case studies: the "verb-raising" construction in Dutch (in chapter 6 "Cross-Serial Dependencies in Dutch") and gapping in English and Dutch (in chapter 7 "Gapping and the Order of Constituents"). Part III discusses issues of computation and human grammatical performance. Chapter 8 talks about "combinators", i.e. operations that map functions onto other functions, and grammars and Chapter 9 comments on processing in context, analyzing a specific architecture for a parser. Throughout the book, the investigation is driven by questions on the language-processing system as a computational whole. The last chapter, "The Syntactic Interface" , represents a summary of the architecture of Steedman's theory as a whole, its role in acquisition and performance, and its relation to other theories of grammar. The reader may wonder why a book describing a particular linguistic theory is entitled "The Syntactic Process". The reason is that its author has in view a theory of natural grammar more directly compatible with certain syntactic phenomena flagrantly disrupting order and constituency and with psychological and computational factors able to map such surface forms onto interpretable meaning representations. Steedman's work and theory can be perceived as an attempt to offer solutions to many puzzling questions that human language has been raising for centuries. The analyses he presents represent a real challenge for those interested in linguistics, covering a wide range of works in the field and opening new territories to be explored.

    References Gazdar, Gerald. 1981. "Unbounded Dependencies and Coordinate Structure". Linguistic Inquiry, 12, 155-184 Joshi, Aravind, Leon Levy and M. Takahashi. 1975. "Tree-adjunct Grammars". Journal of Computer Systems Science, 10, 136-163 Pollard, Carl. 1984. Generalized phrase Structure Grammars, Head Grammars, and Natural Languages. Ph.D. thesis, Stanford University Wilson, Rob and Keil, Frank. 1999. MIT Encyclopedia of Cognitive Sciences