Aspekte van 'n grafeem-na-foneemomsetter vir Afrikaans: ervaring in die African Speech Technology-projek

Abstract

Aspects of a grapheme-to-phoneme converter for Afrikaans: experience in the African Speech Technology project. The main purpose of this article is to describe the contribution of the phonetician/phonologist as team member in the DACST funded project of the consortium African Speech Technology: Promoting the development of the official languages of South Africa through language and speech technology applications. The focus is on the development of a grapheme-to-phoneme converter within PATANA (PATtern ANAlysis), AST's enabling technological tool. A description of the most important constructs is followed by the different steps towards the final product. Three rule-based modules are necessary for a proper functioning of an Afrikaans grapheme-to-phoneme converter: syllable boundaries insertion rules, stress placement rules, and rules translating graphemes to (deterministic) phonemes. The development thereof will be covered, as well as other, secondary sections of PATANA for Afrikaans, such as building a lexicon. Special reference is made to the way in which the phonetic annotations were done, as well as to some problematic aspects thereof. Finally, the relative value of the sources developed in this project will be evaluated, by concentrating on the cooperation between linguist and engineer. A brief focus on the benefits as well as negative factors of this collaboration concludes the article. In hierdie artikel word die ontwikkeling van 'n grafeem-na-foneemomsetter vir Afrikaans beskryf. Dit het geskied in die DACST-befondste projek African Speech Technology: Promoting the development of the official languages of South Africa through language and speech technology applications. Die belangrikste konstrukte waarna in die artikel verwys word, word eers uiteengesit, waarna die verskillende stappe wat gevolg is, verduidelik word. PATANA (PATtern ANAlysis), 'n AST-ontwikkelde instrument, is hier gebruik. Dit maak voorsiening vir verskillende reëlgebaseerde modules. Die reëls wat vir die omsetting ontwikkel is, word verduidelik. Afrikaans benodig drie verskillende tipes reëls, naamlik lettergreepverdelingsreëls, klemtoontoekenningsreëls, en grafeemvervangingsreëls. Hierdie drie tipes reëls word elk ondergebring in een module, wat in genoemde volgorde funksioneer. Daar word ook verwys na die rol van 'n Afrikaanse leksikon, asook die wyse waarop fonetiese transkripsies gedoen is. Problematiese aspekte word belig. Ten slotte word gekyk na die relatiewe waarde van die produk hiervan as bron vir linguistiese navorsing. Die samewerking tussen ingenieur en linguis in hierdie projek word ten slotte geëvalueer.