The text analysis and processing of Thai language text to speech conversion system

Abstract

Summary form only given. The text to speech conversion system is divided into front-end and back-end analysis module speech synthesis module, text analysis and processing will directly determine the effect of naturalness of synthesized speech. In the real Thai texts, there are often contained non-Thai standard characters such as numbers, abbreviations, currency symbols, and so on. Text normalization is the process which is switched the non-standard words into a standard word process. Thai text normalization difficulties mainly expressed for two indexes: The first is the non-standard word recognition judgment; the second is the ambiguity of the non-standard word processing carried disambiguation. This paper proposes a rule-based Thai text and context to determine the combination of the words associated with normalization method. The experimental results show that this method can achieve the desired results, the within test set's accuracy rate is 97.83%, the outside set's test accuracy rate is 97.61%.