Improving the Quality of Speech Synthesis Using Semi-Syllabic Synthesis

This paper describes how to improve the quality of synthesized speech using semi-syllables. The transition to speech synthesis using semi-syllable speech fragments is substantiated. For speech synthesis, the Yakut language from the turkic group of languages is considered. The full base of syllables and semi-syllables is determined with the assistance of a machine newspaper corpora. Narration text is created using the base of semi-syllables for all received Yakut words. With the assistance of the narrator's text, material is collected for the speech corpora of the Yakut language. The speech corpora is divided into words and semi-syllables, and is divided into semi-syllable parts of speech, taking into account word stress. With the assistance of concatenation, a sound file is created for any word in the Yakut language. Measures are given to further improve the quality of speech synthesis.