The broad study of homograph disambiguity for Mandarin speech synthesis

How to increase the intelligibility and naturalness of synthetic speech has drawn much attention in Mandarin text-to-speech (TTS) research. This has always been treated as a bottleneck due to their effects on human perception. However, as qualities of synthetic speech increase for syllables, words or phrases, there is also an increasing need to improve the various components of the text processing. One of these desired improvements for Mandarin speech synthesis is the accuracy of the character-to-sound (CTS) process. From the viewpoint of application, the purpose of speech synthesis should be aimed at making the synthetic speech understandable by humans and minimizing the misunderstanding between them. It thus is very important to increase the accuracy of the CTS process. Such a process is designed to predict phonetic pronunciations from a coarse surface text input and the difficulty mainly results from ambiguous homograph characters. We propose an effective analysis method incorporated with linguistic knowledge to resolve homograph ambiguity. The methods we use in the experiments are discriminating lexical association and a tree-based language model. From the experiment results, we can get about 10% more improvement on the average accuracy rate than the traditional maximum frequency guess approach for most ambiguous homograph characters