A review of CALL-based ASR and its potential application for Malay cued Speech learning tool application

Abstract

Growth of speech technologies to facilitate Computer-Assisted Language Learning (CALL)-based Automatic Speech Recognition (ASR) for speech therapy has significant effects on education and language learning. Speech technologies used in language learning typically focusing on the speech analysis, speech recognition and speech synthesis. Speech analysis enables the acoustic samples of the speech signal to be represented either as a waveform or visual spectrogram. Speech recognition converts a recorded speech signal into text, while speech synthesis generates artificial human speech based on text input. Progress in research and application of speech technologies has made possible of mobile and computer applications of a combination of both speech recognition and speech syntheses. Varieties of language practice applications are also available with affordable price for downloads. Even though these apps supported translation from other languages such as Malay to English or vice versa; these apps present challenges for the Speech-Language Pathologies (SLP) to provide customised coaching according to the Malaysian school syllabus. On closer examination, most of these apps are technology-driven rather than pedagogy-driven. Furthermore, synthetic speech is unable to provide SLPs with an accurate and automatic diagnosis of pronunciation errors in the Malay Language. It is questionable whether these apps can handle prosody and native speech more accurately than at present. This paper attempted to review and determine potential improvement actions to support CALL-based ASR for Malay Language literacy in the context of Malay Cued Speech method.