Joint NMT models for text conversion between traditional Mongolian script and cyrillic Mongolian: a comparative study
DOI:
https://doi.org/10.5564/jimdt.v6i1.3758Keywords:
Machine Translation, Transformer, Neural machine translation (NMT), Deep LearningAbstract
The research aims to develop a high-precision machine learning model for translating between Mongolian and Cyrillic scripts. Although Mongolian and Cyrillic scripts are different scripts of the Mongolian language, their sentence structure, number of words in a sentence, and position are the same. This article presents the results of our research on neural machine translation models tested for translating between the two scripts. The Seq2Seq and Transformer models with attention mechanisms were trained using each tokenization method at the character, subword, and word levels. The accomplishments of these models were assessed regarding the conversion quality, computational efficiency, and the ability to handle the unique characteristics of the two scripts, and the benefits and drawbacks of each model were summarized. It is believed that this comparison helps select the most suitable NMT model for similar tasks. The developed model has a vast potential for development and application as a system for translating between scripts, and it will facilitate some clerical work in public and private sector organizations at all levels providing services in Cyrillic and Mongolian scripts.
Downloads
146
References
[1] I. Sutskever, O. Vinyals, and Q. V. Le, “Sequence-to-Sequence Learning with Neural Networks,” Advances in Neural Information Processing Systems, Vol. 27, Curran Associates Inc., 2014.
[2] Proceedings of the 27th International Conference on Neural Information Processing Systems, Curran Associates Inc., 2014.
[3] D. Bahdanau, K. Cho, and Y. Bengio, “Neural Machine Translation by Jointly Learning to Align and Translate,” arXiv preprint arXiv:1409.0473, 2015.
[4] A. Vaswani, N. Shazeer, N. Parmar, J. Uszkoreit, L. Jones, A. N. Gomez, L. Kaiser, and I. Polosukhin, “Attention Is All You Need,” Advances in Neural Information Processing Systems, edited by I. Guyon, U. V. Luxburg, S. Bengio, H. Wallach, R. Fergus, S. Vishwanathan, R. Garnett, and D. Precup, Curran Associates Inc., 2017, pp. 5998–6008.
[5] R. Sennrich, B. Haddow, and A. Birch, “Neural machine translation of rare words with subword units,” In Proceedings of the 54th Annual Meeting of the Association for Computational Linguistics, pp. 1715–1725, 2016. https://doi.org/10.18653/v1/P16-1162
[6] A. Mao, et al., “Cross-Entropy Loss Functions: Theoretical Analysis and Applications,” In International Conference on Machine Learning, 2023, pp. 23803–23828.
Downloads
Published
How to Cite
Issue
Section
License
Copyright (c) 2024 Uuganbaatar Dulamragchaa
This work is licensed under a Creative Commons Attribution-NonCommercial 4.0 International License.
The authors grant the Journal of Institute of Mathemathics and Digital Technology a license to publish the article and identify itself as the original publisher.
Articles in the Journal of Institute of Mathemathics and Digital Technology are Open Access articles published under a Creative Commons Attribution-NonCommercial 4.0 International License - CC BY NC.
This license permits NonComericial use, distribution and reproduction in any medium, provided the original work is properly cited.
