Steven Bird, John Coleman, Janet Pierrehumbert and James Scobbie
University of Edinburgh, AT&T Bell Laboratories,
Northwestern University, Stanford University
Declarative phonology is a program of research that was motivated in part by the need for theories of phonology that can be implemented on a computer. While it is clear that such a development would be beneficial for both theoretical and field phonology, it is not immediately obvious how one should go about implementing phonological models. The so-called ‘declarative’ approach draws on a key insight from theoretical computer science, where there has been a long tradition of distinguishing between the declaration of a problem and a procedure which computes the solution to that problem. Paradoxically, the kind of problem specifications that are frequently the most useful for computational implementation are those which make the fewest procedural commitments.
The declarative phonology programme is, at its heart, an attempt to do away with the ordered derivations and the concomitant feature-changing rules of traditional generative phonology. In this respect, declarative phonology ties in with some recent developments in theoretical phonology where feature-changing rules have been criticized or explicitly avoided (Rice, 1989; McCarthy, 1991). However, it is also possible to find precedents in the literature on American Structuralist phonemic phonology (Hockett, 1954), Firthian Prosodic Phonology, Natural Generative Phonology (Hooper, 1976; Hudson, 1980) and Montague Phonology (Wheeler, 1981; Bach, 1983). More recently, ‘harmonic’ approaches to phonology arising from work in connectionism (Smolensky, 1986) have also questioned the procedural paradigm but from a perspective which does not clearly differentiate the declaration of grammar from the means of its implementation (Goldsmith, ta; Prince & Smolensky, 1992). Despite this difference, the declarative and connectionist approaches are alike as regards their incorporation of various kinds of constraint satisfaction.
With increasing interest in the interaction between phonology and syntax being expressed in the literature, declarative phonology has something to contribute here too. Constraint-based grammar frameworks such as HPSG (Pollard & Sag, 1987) (manifesting good linguistic coverage and attractive computational properties) have the same metatheoretical commitments as declarative phonology. The prospect for having a computational theory of phonology that is fully integrated with a computational theory of syntax and semantics is now imminent.
A final area of concern is the phonology-phonetics interface. In the declarative framework it makes sense to view the relationship between phonology and phonetics as being one of denotation. Under this view, phonological representations are descriptions of phonetic reality and a particular phonological construct is said to denote a phonetic event (Bird & Klein, 1990; Pierrehumbert, 1990; Scobbie, 1991a; Coleman, 1992).
This article consists of four sections, where each section has been contributed by a different author. The first three sections present reanalyses of phenomena that have previously been thought to require the ability to destructively modify phonological structures. In 2, James Scobbie discusses syllabification in Tashlhiyt Berber and presents a declarative analysis couched in a feature-structure based framework. In 3, Steven Bird investigates vowel harmony in Monta˜nes Spanish and consonant harmony in Chumash and proposes a non-feature-changing account using a finite-state model. In 4, John Coleman presents a brief overview of his reconstruction of Lexical Phonology in a declarative framework. The final section contains a commentary by Janet Pierrehumbert, discussing the achievements and prospects of declarative phonology in relation to generative phonology, Lexical Phonology and laboratory phonology.
2 CONSTRAINT CONFLICT (James M. Scobbie)
2.1 The Phonotactic: General Tendency or Hard Constraint?
It is well-known that rewrite rules fail to capture some generalisations about the level of representation they derive (Kisseberth, 1970; Shibatani, 1973). Defining the well-formed structures of that level using phonotactics enables those patterns to be addressed. Moreover, insofar as the patterns that exist in a language trigger its alternations, the alternations are explained.
If well-formedness constraints are used, it is necessary to decide whether or not to use rewrite rules also. When a grammar employs both formal techniques, their interaction is necessarily an area of concern (Scobbie, 1991b). Some work (e.g. (Singh, 1987; Paradis, 1988)) replaces structural descriptions with phonotactics and structural changes with repair-strategies. Whenever a structure known to be ill-formed at the surface level of representation can be generated during the derivation, it is indeed generated, only to be destructively modified. Therefore one can state general tendencies of distribution directly in the grammar and ‘repair’ those forms generated by the tendency which happen to be in conflict with empirical considerations. Though in these theories phonological representations are intended to be models of aspects of competence, the derivation and the intermediate forms are uninterpreted aspects of the theory. Such hidden elements imbue the theory with greater abstractness, and they decrease the modularity of the theory with respect to the procedures that can be employed to implement it.
Another line of research is to use only constraints acting in consort to describe a level of
representation. If the constraints are broad-stroke general tendencies (such as a syllable’s
disdain for a coda or love of an onset) they will of course sometimes clash in their demands.
Some means must be found of resolving such inconsistencies.
We can avoid an inconsistent grammar by using formal statements of distribution which
fail to clash by virtue of their precision, and by using familiar conventions such as the
Elsewhere Condition. Formalising the universal tendencies with an appropriate amount
of detail dispels constraint conflict. The interaction of these hard constraints is therefore
declarative and compositional. This is the approach advocated here.
Other approaches adopt optimisation techniques which provide a metric capable of
determining the best-formed structures possible in the contradictory circumstances. The
optimal solution the one in which the fewest important constraints are violated. Tendencies
are in fact soft constraints in these theories and are carefully prioritised in a derivational
architecture familiar from connectionism.
In the next section I examine data which has been argued to be ideally suited to
optimisation. I will show that once the tendencies are formalised, they include enough
detail to allow them to be implemented as hard constraints in a derivationally neutral,
declarative way.
(to read to whole paper, please click on the title as Declarative Phonology)
Go to FJU-Lin. and see the page
papers about declarative phonology
key aspects of declarative phonology
declarative phonology
recommanded books:
declarative phonology in Phonology
作者：Charles W. Kreidler
declarative phonology in Asymmetry in Grammar
作者：Anne-Marie Di Sciullo
declarative phonology in Linguistics Today: Facing a Greater Challenge
作者：Piet Van Sterkenburg, International Congress of Linguists
declarative phonology in Phonetics, Phonology, and Cognition
作者：Jacques Durand, Bernard Laks
沒有留言:
張貼留言