INCORPORATING THE KHOREZM DIALECT INTO UZBEK ASR: AN ANALYTICAL APPROACH TO THE EVOLUTION OF SPEECH RECOGNITION SYSTEMS AND RESOURCES

Kurbonova R. U.


Annotatsiya

This article analyzes the developmental stages of automatic speech recognition (ASR) systems and explores the issues of their integration into Uzbek dialects, specifically focusing on the Khorezm dialect. Initially, the global evolution of ASR technologies is examined, ranging from statistical approaches to modern neural network architectures such as Wav2Vec 2.0, Conformer, Whisper, and Samba-ASR. Furthermore, the article analyzes existing ASR resources developed for the standard literary Uzbek language (USC, UzBERT, UzLM, FeruzaSpeech, etc.) and acknowledges the achievements in this domain. Simultaneously, it highlights the characteristics of the Uzbek language as a low-resource language and explicitly addresses the lack of data in open speech corpora regarding the Khorezm dialect, which significantly differs from the standard language both phonetically and lexically. The analysis results indicate that integrating the Khorezm dialect into these systems is crucial for the comprehensive and accurate performance of modern ASR architectures in the Uzbek language. The article substantiates that the most pressing task in the field is the future development of dedicated parallel corpora for the Khorezm dialect, alongside fine-tuning existing advanced neural models based on this dialectal data.


Adabiyotlar

  1. A. Graves, S. Fernández, F. Gomez, and J. Schmidhuber, “Connectionist Temporal Classification: Labelling Unsegmented Sequence Data with Recurrent Neural Networks,” in Proceedings of the 23rd International Conference on Machine Learning (ICML), ACM, 2006, pp. 369–376. doi: 10.1145/1143844.1143891.
  2. A. Baevski, H. Zhou, A. Mohamed, and M. Auli, “wav2vec 2.0: A Framework for Self-Supervised Learning of Speech Representations,” in Advances in Neural Information Processing Systems (NeurIPS), 2020, pp. 12449–12460.
  3. A. Conneau, A. Baevski, R. Collobert, A. Mohamed, and M. Auli, “Unsupervised Cross-Lingual Representation Learning for Speech Recognition,” in Proceedings of Interspeech 2021, 2021, pp. 2426–2430. doi: 10.21437/Interspeech.2021-329.
  4. A. Radford, J. W. Kim, T. Xu, G. Brockman, C. McLeavey, and I. Sutskever, “Robust Speech Recognition via Large-Scale Weak Supervision,” in Proceedings of the 40th International Conference on Machine Learning (ICML), 2023, pp. 28492–28518.
  5. M. Musaev, I. Khujayorov, and M. Ochilov, “USC: An Open-Source Uzbek Speech Corpus and Initial Speech Recognition Experiments,” in Speech and Computer (SPECOM 2021), in Lecture Notes in Computer Science, vol. 12997. Springer, 2021, pp. 437–447. doi: 10.1007/978-3-030-87802-3_40.
  6. A. Mukhamadiyev, I. Khujayarov, O. Djuraev, and J. Cho, “Automatic Speech Recognition Method Based on Deep Learning Approaches for the Uzbek Language,” Sensors, vol. 22, no. 10, p. 3683, 2022, doi: 10.3390/s22103683.
  7. B. Mansurov and A. Mansurov, “UzBERT: pretraining a BERT model for Uzbek,” 2021. [Online]. Available: https://huggingface.co/coppercitylabs/
  8. A. Povey and K. Povey, “FeruzaSpeech: A 60 Hour Uzbek Read Speech Corpus with Punctuation, Casing, and Context,” Apr. 2024. doi: 10.48550/arXiv.2410.00035.
  9. A. Ali et al., “The MGB-2 Challenge: Arabic Multi-Dialect Broadcast Media Recognition,” in 2016 IEEE Spoken Language Technology Workshop (SLT), IEEE, 2016, pp. 279–284. doi: 10.1109/SLT.2016.7846277.
  10. S. Mussakhojayeva, Y. Khassanov, and H. A. Varol, “Multilingual End-to-End ASR for Low-Resource Turkic Languages with Common Alphabets,” Sci. Rep., vol. 14, p. 13632, 2024, doi: 10.1038/s41598-024-64848-1.
  11. A. G. McCollum and S. Chen, “Uzbek,” J. Int. Phon. Assoc., vol. 51, no. 3, pp. 1–25, 2021, doi: 10.1017/S0025100319000264.
  12. Z. Dosimov and S. Alimova, “The Vocabulary Layer of the Khorezm Kipchak Dialect,” Psychology and Education Journal, vol. 58, no. 1, 2021.
  13. M. Sharipov, J. Mattiev, J. Sobirov, and R. Baltayev, “Creating a morphological and syntactic tagged corpus for the Uzbek language,” arXiv preprint arXiv:2210.15234, 2022.