4-санға «Автоматтандыру және ақпараттық технологиялар», «Азық-түлік инженериясы және биотехнология», сондай-ақ «Химиялық технология» бағыттары бойынша мақалалар қабылдау жабылды.

Приём статей в четвёртый номер по направлениям «Автоматизация и информационные технологии», «Пищевая инженерия и биотехнология», а также «Химическая технология» закрыт.

Submissions for the fourth issue in the fields of Automation and Information Technology, Food Engineering and Biotechnology, as well as Chemical Technology, are now closed.

Preview

Bulletin of Shakarim University. Technical Sciences

Bulletin of Shakarim University. Technical Sciences

Advanced search

DEEP LEARNING APPROACHES FOR CYBERBULLYING DETECTION IN MULTILINGUAL ONLINE TEXTS

https://doi.org/10.53360/2788-7995-2026-3(23)-3

Abstract

This study presents a systematic comparative analysis of the effectiveness of classical machine learning algorithms and modern deep learning architectures for automatic cyberbullying detection in multilingual online texts (Russian and Kazakh). Particular attention is given to the comparison of traditional feature-based approaches (TF-IDF, n-grams) and models utilizing contextual text representations. For the Kazakh-language corpus, a specialized dataset was developed and manually annotated to ensure high-quality labeling under low-resource language conditions and limited training data availability.
Experimental results showed that the attention-based model achieved the best performance on the Russian-language corpus, reaching an Accuracy of 92.1%, an F1-score of 92.9%, and a ROC-AUC of 0.95. The same architecture also demonstrated the highest results on the Kazakh-language corpus, achieving an Accuracy of 91.5%, an F1-score of 92.9%, and a ROC-AUC of 0.94. Recurrent neural network models, including LSTM and BiLSTM, showed stable performance within the range of 90-91%, whereas classical machine learning algorithms yielded lower results across all major evaluation metrics. These findings confirm the effectiveness of deep learning approaches for cyberbullying detection in multilingual textual environments.
The study highlights the importance of integrating multimodal and interpretable approaches, expanding linguistic resources, and applying transfer learning along with pre-trained language models to improve robustness and classification accuracy. The obtained results have practical significance for intelligent content moderation systems, digital safety applications, and the early detection of psychological risk factors among users in online environments.

About the Authors

R. B. Abdrakhmanov
International University of Tourism and Hospitality
Kazakhstan

Rustam Abdrakhmanov – Candidate of Technical Sciences, Associate Professor of the Humanities school

161200, Republic of Kazakhstan, Turkestan, 14A Rabiga Sultan Begim Street 



A. S. Tolep
Khoja Akhmet Yassawi International Kazakh-Turkish University
Kazakhstan

Abdimukhan Tolep – Candidate of Technical Sciences, Associate Professor

161200, Republic of Kazakhstan, Turkestan, 29 Bekzat Sattarkhanov Avenue 



D. R. Sultan
Narxoz University
Kazakhstan

Daniyar Sultan – PhD, Ac. Associate Professor of the School of Digital Technologies

050035, Республика Казахстан, г. Алматы, ул. Жандосова, 55



A. S. Kasymbekov
International University of Tourism and Hospitality
Kazakhstan

Abbazzbek Kasymbekov – Candidate of Technical Sciences, Associate Professor of the Humanities school

161200, Republic of Kazakhstan, Turkestan, 14A Rabiga Sultan Begim Street



A. B. Nurzhanova
L.N. Gumlyov Eurasian National University
Kazakhstan

Aizhan Nurzhanova – Doctoral student

010008, Republic of Kazakhstan, Astana, 2 Satpayev Street



References

1. Mahmud T.S.D.A. Exhaustive Study into Machine Learning and Deep Learning Methods for Multilingual Cyberbullying Detection in Bangla and Chittagonian Texts / T.S.D.A. Mahmud, M. Ptaszynski, F. Masui // Electronics. – 2024. https://doi.org/10.3390/electronics13091677.

2. Tahir S. HuEID: Hybrid Deep Learning for Cyberbullying Detection using Multi-Modal Urdu Text and Emojis / S. Tahir, A. Nawaz. – 2025. https://doi.org/10.1145/3769294.

3. Khafajeh H. Cyberbullying Detection in Social Networks Using Deep Learning / H. Khafajeh // The International Arab Journal of Information Technology. – 2024. – № 21(6). https://doi.org/10.34028/iajit/21/6/9.

4. Mahajan E. EnsMulHateCyb: Multilingual hate speech and cyberbully detection in online social media / E. Mahajan, H. Mahajan, S. Kumar // Expert Systems With Applications. – 2023. – № 236. – Р. 121228. https://doi.org/10.1016/j.eswa.2023.121228.

5. Begum A. Cyberbullying Detection on Social Media using AI / A. Begum // International Journal For Science Technology And Engineering. – 2025. – № 13(4). – Р. 7003-7010. https://doi.org/10.22214/ijraset.2025.70045.

6. Musyoka I. Multimodal Cyberbullying Detection Using Deep Learning Techniques: A Review / I. Musyoka, J. Wandeto, B. Kituku. – 2023. – Р. 187-192. https://doi.org/10.1109/ict4da59526.2023.10302244.

7. Deep Learning for Multi-Labeled Cyberbully Detection / N. Islam et al // Enhancing Online Safety. – 2023. – Р. 1-6. https://doi.org/10.1109/icdsns58469.2023.10245135.

8. Vijayakumar H.P.D.D. A study on deep learning algorithms for multimodal and multilingual cyberbullying detection / H.P.D.D. Vijayakumar // International Journal of Scientific Research. – 2021. – Р. 50-54. https://doi.org/10.36106/IJAR/4620274.

9. Horate O. Cyber-Bullying and Harassment Detection using ML / O. Horate // International Journal for Science Technology and Engineering. – 2025. – № 13(4). – Р. 5761-5766. https://doi.org/10.22214/ijraset.2025.69732.

10. AI-Powered Frameworks for the Detection and Prevention of Cyberbullying Across Social Media Ecosystems / M.A.M. Mondol et al // Techcomp Innovations. – 2025. – № 2(1). – Р. 1-15. https://doi.org/10.70063/techcompinnovations.v2i1.59.


Review

For citations:


Abdrakhmanov R.B., Tolep A.S., Sultan D.R., Kasymbekov A.S., Nurzhanova A.B. DEEP LEARNING APPROACHES FOR CYBERBULLYING DETECTION IN MULTILINGUAL ONLINE TEXTS. Bulletin of Shakarim University. Technical Sciences. 2026;(3(23)):23-30. (In Russ.) https://doi.org/10.53360/2788-7995-2026-3(23)-3

Views: 16

JATS XML


Creative Commons License
This work is licensed under a Creative Commons Attribution 4.0 License.


ISSN 2788-7995 (Print)
ISSN 3006-0524 (Online)
X