APPLICATION OF BLEU AND SARI METRICS IN EVALUATING SIMPLIFIED TEXTS IN KAZAKH: ANALYSIS AND EFFECTIVENESS

This article explores the methodology for evaluating the quality of simplified texts in Kazakh using BLEU and SARI metrics. Text simplification is an important aspect for ensuring information accessibility and facilitating the learning process in Kazakh language. The BLEU metric, based on comparing n-grams of translation and reference, is widely used for evaluating the quality of machine translation, but it does not take the context of the input text into account. The SARI metric, specifically designed for evaluating text simplification, considers semantic changes and shows a higher correlation with human judgments. In this study, algorithms for replacing complex words with simple synonyms and for replacing or removing complex phrases were applied. The analysis results showed that the SARI metric is more sensitive to the changes made in simplified texts compared to BLEU. Therefore, the combined use of BLEU and SARI metrics provides a comprehensive and accurate evaluation of the quality of simplified texts in Kazakh.

1. Jiang, Chao, Mounica Maddela, Wuwei Lan, Yang Zhong, Wei Xu. Comput. Res. Repos., 2020. https://doi.org/10.48550/arXiv.2005.02324

2. Lindstrom, Jennifer H. Teaching Exceptional Children, 2019, vol. 51, pp. 189–200. https://doi.org/10.1177/0040059918763712.

3. Xu W., Callison-Burch C., Napoles C. Transactions of the Association for Computational Linguistics, 2015, vol. 3, pp. 283–297. https://doi.org/10.1162/tacl_a_00139.

4. Suha S. Al-Thanyyan, Aqil M. Azmi. ACM Comput. Surv., 2021, 36 p. https://doi.org/10.1145/3442695.

5. Papineni K., Roukos S., Ward T., Zhu W.-J. Proceedings of the 40th Annual Meeting on Association for Computational Linguistics – ACL ’02, 2001, p. 311. https://doi.org/10.3115/1073083.1073135.

6. Xu, Wei, Napoles, Courtney, Pavlick, Ellie, Chen, Quanze, Callison-Burch, Chris. Transactions of the Association for Computational Linguistics, 2016, pp. 401–415. https://doi.org/10.1162/tacl_a_00107.

7. Narayan S., Gardent C. Proceedings of the 52nd Annual Meeting of the Association for Computational Linguistics, 2014, vol. 1, pp. 435–445. https://doi.org/10.3115/v1/P14-1041.

8. Sulem E., Abend O., Rappoport A. Proceedings of the 2018 Conference on Empirical Methods in Natural Language Processing, 2018, pp. 738–744. https://doi.org/10.18653/v1/D18-1081.

9. 1Sulem E., Abend O., Rappoport A. Proceedings of the 2018 Conference of the North American Chapter of the Association for Computational Linguistics: Human Language Technologies, 2018, vol. 1, pp. 685–696. https://doi.org/10.48550/arXiv.1810.05022.

10. Qonaqjailyqqa negızdelgen qazaq ūlttyq ashanasy, soyle.kz, 2024. Available: https://www.soyle.kz/article/view?id=879. [Accessed: 26-Nov- 2024] [in Kazakh]

11. 1Janfada B., Minaei-Bidgoli B. 6th International Conference on Web Research (ICWR), 2020, p. 271. https://doi.org/10.1109/ICWR49608.2020.9122325.