The implementation of orthoepic norms in the Kazakh speech synthesis system

Abstract

The article discusses the issues of modeling the speech synthesis system in the Kazakh language based on orthoepic norms. As a result of the rapid development of digital technologies, the importance of speech synthesis systems aimed at the natural-sounding of text in linguistics is increasing. In this context, the article analyzes the scientific foundations of the use of phonetic, orthographic, and prosodic regularities of the Kazakh language in the speech synthesis system. In particular, the importance of maintaining orthographic norms when converting written text into spoken speech is highlighted. The quality of a speech synthesis system is determined not only by the grammatically correct processing of the text, but also by its natural and literary sounding. Accordingly, sound interaction and the accurate representation of intonational characteristics are essential for producing naturalsounding synthesized speech. Ignoring orthoepic norms can reduce the naturalness of speech and negatively affect listeners’ perception and comprehension. In this regard, the introduction of the sound patterns of the Kazakh language into the speech synthesis system is one of the current scientific issues. The paper highlights the history of studying orthoepy in Kazakh linguistics and current research trends, as well as reviews both domestic and international studies. Based on the analysis, the necessity of systematically modeling the phoneticphonological and prosodic features of text to improve the quality of synthesized speech is emphasized. The research draws upon the works of scholars such as I. Kenesbayev, M. Balakayev, S. Myrzabekov, R. Syzdyk, A. Zhunisbek, and Z. Bazarbayeva, while theoretical issues of speech synthesis are considered in comparison with international studies. In addition, the characteristics of modern Kazakh speech synthesis systems (Kazakh TTS2, Kazakh ASR, and other projects) are analyzed, and the level of orthoepic norm implementation in these systems is described. The study considers major speech synthesis approaches, including compiler-based, parametric, formant-based, and neural network methods, and examines their applicability to the modeling of linguistic material. The necessity of considering phonemic changes, positional and combinatorial features of sounds, intonational structure, and rhythmic groups in speech synthesis systems is emphasized. Based on the conducted research, the systematic use of an orthoepic algorithm and prosodic annotation is proposed for improving the Kazakh speech synthesis system. The article also analyzes the adaptation of loanwords from an orthoepic perspective and outlines ways of their phonetic integration in accordance with the phonological laws of the Kazakh language. The results of the study demonstrate that systematic modeling of orthoepic norms and prosodic features enhances the naturalness of synthesized speech and facilitates listener comprehension. Furthermore, the development of the Kazakh speech synthesis system contributes to the broader use of the national language in the digital environment.