Arabic Discourse Analysis: A Naïve Algorithm for Defective Pronunciation Correction

Abstract

This manuscript summarizes a novel study to detect and correct voice anomalies contained in Arabic discourses. These objectives are attained by following some fundamental steps. The first consists in classifying the Arabic produced healthy or pathological vocal signals. Second, the identification of problematic phonemes takes place. The proposition of algorithm allowing the correction of defective pronunciations presents the aim of the latter task. We are satisfied with the obtained results. In deed, the elaborated algorithm has attained a correction performance of 90% based on 52 Arabic voice sequences covering male and female, healthy and pathological speeches and clustering several areas. Consequently, researchers of topics related to speech processing can benefit from our proposition in the conception and development of their systems.