Application of Amharic Speech Recognition System for Dictation

Prototype of a Dictation application for Amharic language is implemented and tested with different, techniques, methodologies and tools inside the contents of this book. This book can be used, for different researchers and developers that are interested in working speech based applications for any language, as a very good guide. Sphinx packages are used for modeling and developing the major components of the speech recognizer system. It shows how to integrate open source word processing application open office with sphinx-4 decoder and how to build windows application with java that can be used as an rtf text editor and can be used as a dictation prototype by integrating the app with sphinx-4 core decoder with xml configuration. During modeling of the recognizer core components, Acoustic model, Language Model and lexical(pronunciation dictionary) model, hmm(hidden markov model) is applied for acoustic model production, tri-gram language modeling method was applied for language model production and canonical pronunciation dictionary with basic Amharic language phones were prepared with python scripts specifically coded for this purpose.