This paper describes the creation of a prototype of the Chechen speech synthesis system that is based on the DCTTS neural network model and consists of various functional modules. It analyzes the quality of the received speech signal in accordance with the MOS international assessment of the quality of synthesized speech, as well as the results of testing the prototype of the Chechen speech synthesis system and its refinement. New areas of research are presented, which are preliminarily focused on the elimination of graphic homonymy in the text.
Automatic Documentation and Mathematical Linguistics – Springer Journals
Published: Jul 12, 2021
Keywords: Chechen speech synthesis system; training corpus; analysis of the quality of synthesized speech; MOS assessment; graphic homonymy