CREATION OF ARTIFICIAL INTELLIGENCE-BASED CONTINUOUS SPEECH RECOGNITION SYSTEM FOR UZBEKISTAN: DESIGN OF CORPUS, ACOUSTIC MODEL AND LANGUAGE MODEL
DOI:
https://doi.org/10.47390/ts-v4i3y2026N04Keywords:
artificial intelligence; Uzbek language; automatic speech recognition; speech corpus; acoustic model; language model; deep learning; CTC-attention; WER; CER.Abstract
This article analyzes the scientific basis for creating a continuous speech recognition system based on artificial intelligence for the Uzbek language. The study systematically summarizes the stages of corpus formation, data preprocessing, acoustic modeling, language model construction, decoding and results evaluation based on the primary sources under investigation. Comparative analysis, structural synthesis and interpretation of published results were used as a methodological basis. The studied literature shows that the quality of speech recognition in the Uzbek language is determined, first of all, by the quality and size of the corpus, the preprocessing discipline, hybrid deep learning architectures and language models adapted to the agglutinative nature of the Uzbek language. Based on these results, a six-stage architecture is proposed, consisting of corpus formation, preprocessing, character separation, acoustic modeling, language model construction and decoding with error evaluation.
References
1. Russell, S., & Norvig, P. (2022). Artificial Intelligence: A Modern Approach (4th ed.). Pearson.
2. Goodfellow, I., Bengio, Y., & Courville, A. (2016). Deep Learning. MIT Press.
3. Bishop, C. M. (2006). Pattern Recognition and Machine Learning. Springer.
4. Murphy, K. P. (2022). Probabilistic Machine Learning: An Introduction. MIT Press.
5. Musaev, M., Mussakhojayeva, S., Khujayorov, I., Khassanov, Y., Ochilov, M., & Varol, H. A. (2021). USC: An Open-Source Uzbek Speech Corpus and Initial Speech Recognition Experiments. In A. Karpov & R. Potapova (Eds.), Speech and Computer, SPECOM 2021, Lecture Notes in Computer Science, Vol. 12997 (pp. 437–447). Springer. https://doi.org/10.1007/978-3-030-87802-3_40
6. Musaev, M., Khujayorov, I., & Ochilov, M. (2020). Development of integral model of speech recognition system for Uzbek language. In 2020 IEEE 14th International Conference on Application of Information and Communication Technologies (AICT) (pp. 1–6). IEEE. https://doi.org/10.1109/AICT50176.2020.9368719
7. Mukhamadiyev, A., Khujayarov, I., Djuraev, O., & Cho, J. (2022). Automatic Speech Recognition Method Based on Deep Learning Approaches for Uzbek Language. Sensors, 22(10), 3683. https://doi.org/10.3390/s22103683
8. Mukhamadiyev, A., Mukhiddinov, M., Khujayarov, I., Ochilov, M., & Cho, J. (2023). Development of Language Models for Continuous Uzbek Speech Recognition System. Sensors, 23(3), 1145. https://doi.org/10.3390/s23031145
9. Madrakhimov, S., Makharov, K., & Lolaev, M. (2021). Data preprocessing on input. AIP Conference Proceedings, 2365(1), 030003. https://doi.org/10.1063/5.0058132
10. Zaynidinov, H., & Mallayev, O. (2022). Parallel algorithm for calculating the learning processes of an artificial neural network. AIP Conference Proceedings, 2647(1), 050006. https://doi.org/10.1063/5.0104178
11. Musaev, M., Khujayarov, I., & Ochilov, M. (2023). Speech Recognition Technologies Based on Artificial Intelligence Algorithms. In H. Zaynidinov, M. Singh, U. S. Tiwary, & D. Singh (Eds.), Intelligent Human Computer Interaction, IHCI 2022, Lecture Notes in Computer Science, Vol. 13741 (pp. 51–62). Springer. https://doi.org/10.1007/978-3-031-27199-1_6
This work is licensed under a 