Unicode Sinhala and phonetic English bi-directional conversion for Sinhala speech recognizer

An automated speech recognizer (ASR) having a large vocabulary is yet to be developed for the Sinhala language because of the time consumed in gathering the training data to build a language model. The dictionary and building the language model require non-English text, in our case, Sinhala Unicode, to be transcribed in phonetic English text. Unlike text to speech conversions which only require transcribing the non- English text to phonetic English text, an ASR needs correct reproduction of the original language text when the phonetic English text is produced as the output of the speech recognizer. In the present research, newspaper articles are used to gather a large set of sentences to build a language model having thousands of words for the Sphinx ASR. We present a decoder algorithm that produces phonetic English text from Sinhala Unicode text and an encoder algorithm that produces the correct reproduction of Unicode Sinhala text from phonetic English. For a near phonetic tag set for Sinhala alphabet, results indicate 100% accuracy for the decoder algorithm while for numberless text, accuracy of the encoder algorithm stands at 98.61% for distinct phonetic English words.