Creating and Editing Voice Sites with Embedded Speech Recognition System

Abstract

World Wide Telecom Web (WWTW) is a voice-driven ecosystem parallel to WWW. WWTW simply defined as a network of interconnected voice sites that are voice driven applications created by users and hosted in the network. It has proved that it has the potential to join underprivileged population to the next generation converged networked world. Since WWTW can be browsed over speech, it has a voice browser at server side. The core components of any voice browser are ASR (Speech Recognition) engine and TTS (Text To Speech) engine. The most crucial component of these two is ASR engine which is used to recognize subscriber s voice commands which drives the system as subscriber wants. The other way ASR is being used when subscriber needs to create or edit his or her voice site. The ASR engine to recognize subscriber speech and convert it to text which is response to voice template prompt which played by voice browser. Then pass this converted text to voice site creator to create or edit subscriber s voice site. The process of converting speech input to text done at server side. This approach is not good for situation such as concurrent use of WWTW and when user voice input voice is not audible and clear. As a solution to this problem, we proposed to deploy an ASR engine at client device and that ASR engine is only being used when subscribers create or edit their voice sites. In this work we discussed how well this solution suits to the world Wide Telecom Web.