A classifier model for detecting pronunciation errors regarding the Nasa Yuwe language’s 32 vowels

Abstract

The Nasa Yuwe language has 32 oral and nasal vowels thereby leading to one being used instead of the other; such confusion can change the meaning of the spoken word in Nasa Yuwe. A set of classifier models have been developed to support correct learning of this language which is in danger of extinction aimed at detecting confusion in the pronunciation of the 32 vowels; about 85% were obtained after experimenting with a variety of linear and nonlinear classifiers, rates of sensitivity, specificity and accuracy. A support software prototype was designed with these trained classifiers for the correct pronunciation of the language’s vowels.

Keywords

• Classifier model
• Nasa Yuwe language
• pronunciation correction
• pattern recognition
• computer-assisted language learning (CALL)