Spoken Malay Language Influence on Automatic Transcription and Segmentation

The influence of Malay language into modeling a Malay speech lexicon can be potentially useful for a more accurate transcription and segmentation. The problem arises when trying to discriminate the boundaries between similar sounding phonemes for segmentation, especially in dyslexic children‘s speech when reading, which have been influenced by the surrounding phonemes (before and after) thus making it harder to distinguish. Hence, this paper explores the need to model spoken Malay into the read speech lexical model that takes into consideration contextdependent model. By modeling spoken Malay language into the lexical model, better transcription can potentially be achieved with regards to the speech data with highly phonetically similar reading errors.