Computational aids for Zulu natural language processing

Abstract

In this paper the development of two basic computational aids in Zulu natural language processing, namely a morphological analyser, built with the Xerox finite-state tools (Beesley & Karttunen, 2003) and a machine-readable lexicon as an XML document, are discussed. We briefly consider the linguistic characteristics of an agglutinating language such as Zulu, with specific reference to the noun and noun-based words. The issues of computational morphology and the challenges involved in building a morphological analyser for Zulu are addressed, and a brief explanation of how the Xerox finite-state tools may be used for this purpose, is given. Then the development of the morphological analyser and the lexicon are outlined, and finally we discuss the integration and use of these two computational aids to reflect the dynamic nature of natural language by focusing on a variant of the morphological analyser, namely the so-called 'guesser'. By applying this guesser to the available language corpora, new word roots may be identified and systematically included in the XML lexicon and the morphological analyser.