APPLICATION OF SPEED READING TECHIQUES IN TEXT-TO-SPEECH GENERATION

摘要

A method and device for converting text to speech such that playback duration is decreased while the comprehensibility of the generated speech is not significantly reduced is disclosed. A text segment initially undergoes linguistic profiling wherein a playing rate indicator for each word, and optionally each element of punctuation, is determined. The playing rate indicator is set to reflect the importance of the associated word or element of punctuation as ascertained through an application of speed-reading techniques, such as matching against a pre-selected word inventory, grammatical analysis, or punctuation analysis. As well, the playing rate indicator may reflect certain linguistic characteristics of the associated word, such as its length. The text is subsequently converted to speech by a text-to-speech engine capable of varying the playing speed of each word, and each pause associated with punctuation, in the text segment according to the corresponding playing rate indicator of the word or element of punctuation.