<?xml version="1.0" standalone="yes"?>
<Paper uid="J94-3000">
  <Title>amp;quot;.on a -</Title>
  <Section position="4" start_page="0" end_page="0" type="intro">
    <SectionTitle>
3. Research Themes
</SectionTitle>
    <Paragraph position="0"> I have attempted to identify four strands of work and cite a representative sample of work within each. Unfortunately, much valuable, relevant work has had to be omitted from the citation lists below for reasons of space.</Paragraph>
    <Paragraph position="1"> Formal reconstruction and language-theoretic results. Work in this area seeks to provide coherent and well-understood formal frameworks in which phonological theories can be expressed. Some work takes an existing theory as its starting point and seeks to refine it and express it in increasing levels of formality, while other work begins from an existing formalism and tries to adapt its expressive capabilities to the needs of phonology. Since most work contains a mixture of both, I shall not attempt a classification. Rather, I shall loosely classify a selection of the work based on the formal method used: unification (Carson 1988; Chung 1990; Coleman 1991; Scobbie 1991; Broe 1993; Walther 1993), predicate logic (Bird 1990; Bouma 1991; Russell 1993), modal logic (Bird and Blackburn 1991; Calder and Bird 1991), type theory (Klein 1991; Mastroianni 1993), categorial grammar/logic (Wheeler 1981; Dogil 1984; van der Linden 1991; Oehrle 1991; Steedman 1991; Moortgat and Morrill to appear), finite-state devices (Kay 1987; Kornai 1991; Wiebe 1992; Bird and Ellison 1994), electrical circuitry (Gilbers 1992), and formal language theory (Ristad 1990; Kornai 1991; Ritchie 1992; Wiebe 1992).</Paragraph>
    <Paragraph position="2"> V Computational Linguistics Volume 20, Number 3 This work addresses phonological theories such as autosegmental, metrical, underspecification, and government phonology. The paper by Kaplan and Kay in this collection is another example of work in this general vein.</Paragraph>
    <Paragraph position="3"> Implementations. Work in this area is directed at producing computer programs that can be used by phonologists to develop and test theories. A variety of SPE implementations exist (independently of the finite-state transducer model) starting from Bobrow and Fraser (1968) and including models for applying rules in reverse (Bear 1990; Maxwell 1991). Other theoretical frameworks that have been implemented to a greater or lesser extent include lexical phonology (Williams 1991), autosegmental phonology (Bird 1990; Albro 1994; Bird and Ellison 1994), diachronic phonology (Hewson 1974; Eastlack 1977; Lowe and Mazaudon 1989), inheritance-based models (Daelemans 1987; Reinhard and Gibbon 1991) and connectionist models (see the next paragraph on learning). The paper by Lowe and Mazaudon in this collection is an example of other work under the heading of implementations.</Paragraph>
    <Paragraph position="4"> Automatic learning. This work aims to provide models to (i) simulate human behavior and test of theories of human language acquisition, and (ii) provide the working phonologist with useful generalizations about a certain body of data under study. Examples of the first type are (Lathroum 1989; Touretzky and Wheeler 1993; Gupta and Touretzky 1992; Hare 1990; Gasser and Lee 1990; Gasser 1992; Shillcock et al. 1992; Goldsmith 1993; Larson 1992), and these all use connectionist models.</Paragraph>
    <Paragraph position="5"> Examples of the second type are all symbolic (Johnson 1984; Dresher and Kaye 1990; Ellison 1993; Bird 1994). Daelemans, Gillis, and Durieux have contributed a paper to the present collection that fits into this category of automatic learning.</Paragraph>
    <Paragraph position="6"> Interfacing to grammar and speech. The final grouping contains work that is intended to integrate computational models of phonology with computational models of grammar and of speech. Concerning the phonology-grammar interface, all this work is covered under the paragraph on formal reconstruction above. The assumption is that if phonological models are formalized and if they employ the same computational model as is used for computational syntax and semantics, then interfacing to grammar ought to be relatively straightforward. Another instance of this work is the contribution to the present collection by Bird and Klein. Recent work on integrating phonology with speech synthesis includes Hertz (1990), Coleman (1992), and Dirksen (1992), and there is also a large literature on the phonology of intonation as it relates to synthesis (e.g. Anderson, Pierrehumbert, and Liberman 1984; Ladd 1987).</Paragraph>
    <Paragraph position="7"> This concludes the discussion of the various current research themes in computational phonology. As chance would have it, each of these themes is manifested by one of the papers in the present collection. We now go on to survey these papers briefly. The reader is referred to the commentaries for more detailed overviews of the contributions.</Paragraph>
  </Section>
class="xml-element"></Paper>