The multiple pronunciations in Taiwanese and the automatic transcription of Buddhist sutra with augmented read speech

Collection of Taiwanese text corpus with phonetic transcription suffers from the problems of multiple pronunciation, or pronunciation variation. By further augmenting the text with read speech, and using automatic speech recognition with a sausage searching net constructed from the multiple pronunciations of the text corresponding to its speech utterance, we are able to reduce the effort for phonetic transcription. Compared to general method for pronunciation variation such as the relabeling of training corpus of [1], the sausage searching net shows advantages. Two experiments are conducted using a Taiwanese Buddhist Sutra speech and text corpus.