Brief overview

The Delta programming language is designed for formalizing and testing phonological and phonetic theories. Its central data structure lets linguists represent utterances as multiple “streams” of synchronized units of their choice, giving them considerable flexibility in expressing the relationship between phonological and phonetic units. This paper presents Version 2 of the Delta language, showing how it can be applied to two linguistic models, one for Bambara tone and fundamental frequency patterns and one for English formant patterns. While Delta is a powerful, special-purpose language that alone should serve the needs of most phonologists, phoneticians, and linguistics students who wish to test their rules, the Delta System also provides the flexibility of a general-purpose language by letting users intermingle C programming language statements with Delta statements.

Introduction

Despite their common interest in studying the sounds of human language, the fields of phonology and phonetics have developed largely independently in recent years. One of the contributing factors to this unfortunate division has been the lack of linguistic rule development systems. Such systems are needed to let linguists easily express utterance representations and rules, and facilitate the computational implementation and testing of phonological and phonetic models.

SRS (Hertz 1982) is a rule development system that was designed, starting in 1974, for just this purpose – to let linguists easily test phonological and phonetic rules, and explore the interface between phonology and phonetics through speech synthesis. SRS, however, was influenced quite heavily by the theory of generative phonology that was prevalent at the time, a theory that posited linear utterance representations consisting of a sequence of phoneme-sized segments represented as bundles of features (Chomsky and Halle 1968).