Research Of Segmentation Dictionary And Markup Language In Chinese TTS Engine

With the development of computer-related technology, voice interaction has become an effective means of human-computer interaction. Let the computer sound is a natural and smooth thing people dream of, with the development of linguistics, phonetics and computer technology, this goal is getting closer and closer to reality. Speech synthesis (TTS) is the technology which make text message into voice. It can make the computer read the text fluently and understand the content of information. Although in recent years, some products of Speech synthesis come out, but there is a gap between these products and mature products that people expect.This article describes the textual analysis of Chinese TTS engine, focuses on the segmentation dictionary and prosodic markup language. Text analysis is the first module of speech synthesis, speech synthesis engine is beginning to lay a good foundation. The piece of text analysis describes the document structure analysis, text normalization, segmentation ambiguity classification and voice, rhythm analysis and multi-tone word. Papers focuses on the Segmentation dictionary, the word comparative analysis of binary-seek-by-word dictionary, Verbatim binary system, Tree index tree and other mechanisms of several common dictionary, learn their strengths, we design a mechanism to improve the dictionary to establish a more effective and rational index, and make a compare between the improved segmentation dictionary and traditional segmentation dictionary. The expression of the Chinese rhythmic awareness, introduced two sophisticated markup language, on this basis, combined with Chinese characteristics, markup language found in the current deficiency in Chinese, designed a line with the rhythm of Chinese prosodic features markup language, markup language based on the experiment and the natural language to compare.In theory and in practice, papers proved that the improved mechanism of the segmentation dictionary although a little more space consuming, but there is a query efficiency greatly improved. Text based on rhythm markup language, especially more complex sentence structure, processed synthesized speech is also superior in naturalness.