A computational morphological analyser for Northern Sotho deverbative nouns: applying Xerox finite-state software to traditional grammar

Abstract

To create a computational morphological analyser, it is necessary to formalize both morphotactic (word formation) rules and morphophonemic (alternation) rules. This article describes the status of a morphological analyser for Northern Sotho deverbative nouns, based on rules extracted from the best available traditional grammars and implemented using Xerox finite-state technology. The rigor of creating a computational morphological analyser, and testing it on large corpora, has led to improvements in the rules themselves and in the understanding of the phenomena. The linguistic information is based on research involving prominent Northern Sotho grammars and tested against data from electronic written sources. It is highlighted that rigorous computational language formalism elicits interesting results linguistically as linguistic rules need to be re-examined more carefully and verified in practice.