This paper explores the challenges and advancements in Grapheme-to-Phoneme (G2P) conversion for the Uzbek language, a low-resource language in speech technology. G2P conversion is a fundamental component in text-to-speech (TTS), automatic speech recognition (ASR), and other natural language processing (NLP) applications. The accuracy of G2P directly affects speech synthesis quality and intelligibility. The study reviews traditional rule-based, statistical, and modern deep learning approaches to G2P modeling, highlighting their advantages and limitations. Particular attention is paid to the phonological features of Uzbek, such as vowel harmony, loanwords from Russian, Arabic, and Persian, and homographs, which pose specific difficulties in phoneme prediction. Existing tools like Phonetisaurus, Sequitur-G2P, and Muxlisa AI are evaluated, noting their potential and current shortcomings. Muxlisa AI represents a significant step in developing TTS systems for Uzbek, yet its reliance on rule-based methods limits pronunciation accuracy. The paper emphasizes the importance of integrating morphological analysis, prosodic modeling, and the International Phonetic Alphabet (IPA) for improved phoneme alignment and synthesis. Furthermore, it suggests adopting hybrid models combining rule-based systems with neural networks to overcome dataset scarcity and phonological complexity. The findings underscore the need for continued research in Uzbek-specific G2P modeling to improve speech technology solutions for underrepresented languages. By enhancing phonetic precision and naturalness, robust G2P models will significantly contribute to digital inclusivity and the broader application of AI-driven linguistic tools in Uzbek.

HAMROYEVA SH.M.

Doctor of philology, docent, Tashkent State University of Uzbek language and literature named after A. Navoi, Republic Uzbekistan

E-mail: hamroyeva81@mail.ru, https://orcid.org/0000-0002-5429-4708

MAKHMUDJONOVA G.U.

PhD student, Department of Computational Linguistics and Digital Technologies, Tashkent State University of Uzbek Language and Literature named after A. Navoi, Tashkent Uzbekistan.

E-mail: gulshaxnozmahmudjonova@gmail.com, https://orcid.org/0009-0002-8536-0680

1. https://docs.nvidia.com/nemo-framework/user-guide/24.09/nemotoolkit/tts/g2p.html
2. Taylor, P. Text-to-Speech Synthesis. Cambridge University Press, 2009.
3. Yi L., Li J., Hao J., & Xiong Z. Improved Grapheme-to-Phoneme Conversion for Mandarin TTS. Tsinghua Science & Technology, (2009), 14, 606-611. doi: 10.1016/S1007-0214(09)70124-5.
4. Bisani M., & Ney H. Joint-Sequence Models for Grapheme-to-Phoneme Conversion. Speech Communication, (2008).
5. Jiampojamarn S. Grapheme-to-Phoneme Conversion and Its Application to Transliteration. University of Alberta, 2009. Retrieved from https://era.library.ualberta.ca
6. Jolchieva S., Nemet G., & Giresh B. Preobrazovanie grafem v fonemy na osnove transformera. Trudy Interspeech, (2019).
7. Cheng S., Zhu P., Liu J., & Wang Z. A Survey of Grapheme-to-Phoneme Conversion Methods. Applied Sciences, (2024), 14(24), 11790. doi:10.3390/app142411790.
8. Novak J.R., Minematsu N., & Hirose K. WFST-Based Grapheme-to-Phoneme Conversion: Open Source Tools for Alignment, Model-Building and Decoding. Proceedings of the 10th International Workshop on Finite State Methods and Natural Language Processing, Donostia–San Sebastián, Spain, (2012).
9. https://github.com/MontrealCorpusTools/mfa-models
10. https://github.com/AdolfVonKleist/Phonetisaurus
11. https://github.com/rhasspy/piper-phonemize/tree/master
12. https://github.com/lingjzhu/CharsiuG2P
13. https://github.com/sequitur-g2p/sequitur-g2p
14. https://muxlisa.uz
15. Deri A., & Knight K. Grapheme-to-Phoneme Models for (Almost) Any Language. Proceedings of the 54th Annual Meeting of the Association for Computational Linguistics, Berlin, Germany, (2016), 399–408.
16. Mirtozhiev M.M. Fonetika uzbekskogo yazyka. Akademiya nauk Uzbekistana, izdatel'stvo «Fan», Tashkent, (2013).
17. Sharma M. Novel NLP Methods for Improved Text-To-Speech Synthesis (Doctoral dissertation), (2021).
18. https://www.ipachart.com

speech synthesis, phonetics, NLP, text-to-speech (TTS), automatic speech recognition (ASR), uzbek language, grapheme-to-phoneme (G2P)