Empirical measurements on a Sesotho tone labeling algorithm

Abstract

This paper discusses the empirical assessments employed on two versions of a Sesotho tone labeling algorithm. This algorithm uses linguistically-defined Sesotho tonal rules to predict the tone labels on the syllables of Sesotho words. The two versions differ in the number of tonal rules that they employ as well the lexical categories that the tone rules apply to. Both versions were tested on the same input text and we employed t-tests to prove that one version is an improvement on the other.