Machine Learning and its Novel Application to Speech Impediment Therapy

András Kocsor, Dénes Paczolay

A new computer aided software package, the „SpeechMaster” was developed for teaching reading and speech impediment therapy. The program is freely available in the Internet at www.inf.u-szeged.hu/beszedmester. Its substance is a machine learning module, which converts the sounds into easily comprehensive graphical elements such as flickering graphemes. This effect efficiently supports the development of the correct articulation of the hearing impaired. The aim of the reading-teaching part of the software is to help children learn, more quickly and easily, the correct phone-grapheme and grapheme-phone associations using the computer as a means of motivation, and by using automatic speech recognition methods. Besides briefly describing the„SpeechMaster” package, in this paper we also summarize our results from the viewpoint of machine learning.