Linguistic Foundations for Constructing Speech Synthesis in the Kazakh Language

Vollständiger Abstract

Worum geht es in dieser Arbeit?

The article is devoted to the problem of correctly recognizing and pronouncing, in accordance with orthoepic norms, shortened words and abbreviations, numerical symbols, mathematical operations, formulas, and conventional signs within the Kazakh-language text-to-speech system. The main objective of the study is to identify the linguistic features of fully expanding shortened words and abbreviations in texts processed by the speech synthesis system, as well as of representing mathematical expressions in written form, and on this basis to develop a linguistic resource that ensures natural, consistent, and realistic synthesized speech in Kazakh. The article clarifies the distinction between shortened words and abbreviations and proposes several approaches to incorporating them into a speech synthesis program. In particular, it describes methods such as compiling comprehensive lists of shortened forms from domain-specific texts, presenting shortened and full variants in parallel, and preserving internationally established abbreviations in their original forms. The article also provides a dedicated phonetic transcription of shortened words and demonstrates their pronunciation in compliance with orthoepic norms. The article presents model examples of attaching suffixes to shortened words, numerical symbols, and conventional signs in accordance with the vowel harmony rules of the Kazakh language. It is demonstrated that presenting both written and spoken variants of numerical symbols, arithmetic operations, fractions, and complex numbers increases the intelligibility of synthesized text. Furthermore, the parallel use of symbolic representations and established names of chemical elements and formulas is shown to enhance the naturalness of speech synthesis.