Abstract
Last time there are unbelievable results in Natural Language Processing(NLP) and Automatic Speech Recognition(ASR). As a result, everybody can use smart search engines such as ChatGPT, smart voice assistants such as Siri, Alexa and more. But these opportunities are available only to the people who can use English or other common languages. For people who use low-resource languages these products are not available. As collection of transcribed data is time consuming and expensive process, scientists search ways of implementing reliable ASR models for low-resource languages. One of ASR improving methods in the case of lack of data is the use of external language model built on text larger than text in the entire dataset. And use this language model in the decoding process. As Kazakh language is also one of low-resource languages it is was decided to test this approach for kazakh language with different language models like Sequential RNNLM and Transformer LM. Inclusion of language model trained on bigger dataset allowed to decrease error values especially for Word Error Rate (WER). The best result was obtained with Transformer LM, WER was decreased to 7.2%.
Access this chapter
Tax calculation will be finalised at checkout
Purchases are for personal use only
Similar content being viewed by others
References
Ren, Z., Yolwas, N., Slamu, W., Cao, R., Wang, H.: Improving hybrid CTC/attention architecture for agglutinative language speech recognition. Sensors 22, 7319 (2022)
Mamyrbayev, O., Oralbekova, D., Alimhan, K., Nuranbayeva, B.: Hybrid end-to-end model for Kazakh speech recognition. Int. J. Speech Technol. 08, 1–10 (2022)
Kuanyshbay, D., Amirgaliyev, Y., Baimuratov, O.: Development of automatic speech recognition for kazakh language using transfer learning. Int. J. Adv. Trends Comput. Sci. Eng. 9, 5880–5886 (2020)
Mussakhojayeva, S., Dauletbek, K., Yeshpanov, R., Varol, H.A.: Multilingual speech recognition for turkic languages. Information 14(2), 74 (2023). https://doi.org/10.3390/info14020074
Orken, M., Alimhan, K., Oralbekova, D., Bekarystankyzy, A., Zhumazhanov, B.: Identifying the influence of transfer learning method in developing an end-to-end automatic speech recognition system with a low data level. Eastern-Eur. J. Enterp. Technol. 1, 84–92 (2022)
Orken, M., Oralbekova, D., Alimhan, K., Tolganay, T., Othman, M.: A study of transformer-based end-to-end speech recognition system for Kazakh language. Sci. Rep. 12(1), 8337 (2022)
Chuang, S.-P., Liu, A.H., Sung, T.-W., Lee, H.: Improving automatic speech recognition and speech translation via word embedding prediction. IEEE/ACM Trans. Audio Speech Lang. Process. 29, 93–105 (2021). https://doi.org/10.1109/TASLP.2020.3037543
Kubo, Y., Karita, S., Bacchiani, M.: Knowledge transfer from large-scale pretrained language models to end-to-end speech recognizers (2022). https://www.researchgate.net/publication/358655492_Knowledge_Transfer_from_Large-scale_Pretrained_Language_Models_to_End-to-end_Speech_Recognizers
Huang, W.R., Peyser, C., Sainath, T.N., Pang, R., Strohman, T., Kumar, S.: Sentence-select: large-scale language model data selection for rare-word speech recognition. In: Interspeech (2022)
Mukherji, K., Pandharipande, M., Kopparapu, S.K.: Improved language models for ASR using written language text. In: 2022 National Conference on Communications (NCC), Mumbai, India, pp. 362–366 (2022). https://doi.org/10.1109/NCC55593.2022.9806803
Amirgaliyev, Y., Kuanyshbay, D., Yedilkhan, D.: Automatic speech recognition system for Kazakh language using connectionist temporal classifier (2020)
Watanabe, S., et al.: ESPnet: end-to-end speech processing toolkit. In: Proceedings of the Interspeech 2018, pp. 2207–2211 (2018). https://doi.org/10.21437/Interspeech.2018-1456
Watanabe, S., et al.: The 2020 ESPnet Update: new features, broadened applications, performance improvements, and future plans. In: Proceedings of the 2021 IEEE Data Science and Learning Workshop (DSLW) (2021)
Jing, K., Xu, J.: A survey on neural network language models (2019). https://doi.org/10.48550/arXiv.1906.03591
Bengio, Y., Senecal, J.: Quick training of probabilistic neural nets by importance sampling. In: Bishop, Christopher M. and Frey, Brendan J. (eds.) International Conference on Artificial Intelligence and Statistics, Proceedings of the Ninth International Workshop on Artificial Intelligence and Statistics, vol. R4, pp. 17–24 (2003)
Guo, P., et al.: Recent developments on ESPnet toolkit boosted by conformer. In: ICASSP 2021 - 2021 IEEE International Conference on Acoustics, Speech and Signal Processing (ICASSP), Toronto, ON, Canada, pp. 5874–5878 (2021). https://doi.org/10.1109/ICASSP39728.2021.9414858
Bahdanau, D., Cho, K., Bengio, Y.: Neural machine translation by jointly learning to align and translate. ArXiv arXiv:1409.0473 (2014)
Acknowledgement
This research has is funded by the Science Committee of the Ministry of Science and Higher Education of the Republic of Kazakhstan (Grant No. BR11765619).
Author information
Authors and Affiliations
Editor information
Editors and Affiliations
Rights and permissions
Copyright information
© 2023 The Author(s), under exclusive license to Springer Nature Switzerland AG
About this paper
Cite this paper
Bekarystankyzy, A., Mamyrbayev, O., Mendes, M., Oralbekova, D., Zhumazhanov, B., Fazylzhanova, A. (2023). Automatic Speech Recognition Improvement for Kazakh Language with Enhanced Language Model. In: Nguyen, N.T., et al. Recent Challenges in Intelligent Information and Database Systems. ACIIDS 2023. Communications in Computer and Information Science, vol 1863. Springer, Cham. https://doi.org/10.1007/978-3-031-42430-4_44
Download citation
DOI: https://doi.org/10.1007/978-3-031-42430-4_44
Published:
Publisher Name: Springer, Cham
Print ISBN: 978-3-031-42429-8
Online ISBN: 978-3-031-42430-4
eBook Packages: Computer ScienceComputer Science (R0)