BBCS-Net: A BERT-Based CNN-BiLSTM-SVM Hybrid Architecture for Fine-Grained Cyberbullying Detection
Abstract
Cyberbullying detection remains challenging due to diverse linguistic patterns used by offenders. The in- ductive biases of deep learning architectures add to this challenge. Single-model approaches often capture only part of abusive language. This results in distinct but partially overlapping error spaces, limiting their effectiveness in real-world scenarios. To address this issue, we propose BBCS-Net. This is a modular hybrid framework that leverages diverse inductive biases. It combines Convolutional Neural Networks (CNNs), Bidirectional Long Short-Term Memory (BiLSTM) networks, and a margin-based Support Vec- tor Machine (SVM) classifier. In the proposed framework, a Bidirectional Encoder Representations from Transformers (BERT) model is first fine-tuned on a publicly available fine-grained cyberbullying dataset consisting of approximately 47,000 samples across six classes. This generates contextualized word repre- sentations. CNN and BiLSTM models are then trained independently. These models learn complementary feature representations. CNNs capture localized contextual cues. BiLSTMs capture long-range sequential dependencies. Penultimate-layer features from both models are fused at the feature level and classified using an SVM. The proposed approach achieves an average accuracy and F1 Score of about 98.87% us- ing nested cross-validation for model selection. On a held-out test set, it achieves 94.75% accuracy and a macro-F1 score of 94.75%. Error-space analysis shows that BBCS-Net recovers many model-specific errors. It corrects about 70.37% of CNN-only errors and 57.69% of BiLSTM-only errors. This highlights the effectiveness of feature-level hybridization in overcoming inductive-bias-driven failure modes. Latency analysis confirms the framework’s practical feasibility. It achieves a single-sample inference latency of 17.75 ms. During batch processing, the ONNX runtime reduces this to about 4.18 ms per sample. This supports near real-time cyberbullying moderation.References
[1] Mahmud M. I., Mamun M., and Abdelgawad A. (2022) “A Deep Analysis of Textual Features Based Cyberbullying Detection Using Machine Learning,” IEEE Global Conference on Artificial Intelligence and Internet of Things (GCAIoT), IEEE, Alamein New City, Egypt, pp. 166–170. https://doi.org/10.1109/GCAIoT57150.2022.10019058
[2] Mathur S. A., Isarka S., Dharmasivam B., and J. C. D. (2023) “Analysis of Tweets for Cyberbullying Detection,” International Conference on Secure Cyber Computing and Communication (ICSCCC), IEEE, Jalandhar, India, pp. 269–274. https://doi.org/10.1109/ICSCCC58608.2023.10176416
[3] Chinivar S., M. S R., J. S A., and K R V. (2022) “Comparison of Varied Embedding and Machine Learning Classifiers for Fine Grained Offensive Content Identification,” IEEE International Conference for Women in Innovation, Technology & Entrepreneurship (ICWITE), IEEE, Bangalore,
India, pp. 1–6. https://doi.org/10.1109/ICWITE57052.2022.10176223
[4] Chow D. V., Natania F., Nathanael O. T., Setiawan K. E., and Hasani M. F. (2023) “Cyberbullying Detection: An Investigation into NLP and Machine Learning Techniques,” 5th International Conference on Cybernetics and Intelligent System (ICORIS), IEEE, Pangkalpinang, Indonesia, pp. 1–6. https://doi.org/10.1109/ICORIS60118.2023.10352294
[5] Balaji P. G., Katariya P. P., Sruthi S., and Venugopalan M. (2024) “Cyberbullying Detection on Multiclass Data Using Machine Learning and A Hybrid
CNN-BiLSTM Architecture,” ICKECS, IEEE, Chikkaballapur, India, pp. 1–6. https://doi.org/10.1109/ICKECS61492.2024.10616957
[6] Bokolo B. G. and Liu Q. (2023) “Cyberbullying Detection on Social Media Using Machine Learning,” IEEE INFOCOM WKSHPS, IEEE, Hoboken,
NJ, USA, pp. 1–6. https://doi.org/10.1109/INFOCOMWKSHPS57453.2023.10226114
[7] Singh N. K., Singh P., and Chand S. (2022) “Deep Learning Based Methods for Cyberbullying Detection on Social Media,” ICCCIS, IEEE, Greater Noida, India, pp. 521–525. https://doi.org/10.1109/ICCCIS56430.2022.10037729
[8] Joseph V. A., Prathap B. R., and Kumar K. P. (2024) “Detecting Cyberbullying in Twitter: A Multi-Model Approach,” ICDECS, IEEE, Bangalore, India, pp. 1–6. https://doi.org/10.1109/ICDECS59733.2023.10502699
[9] Tapaopong W., Charoenphon A., Raksasri J., and Samanchuen T. (2024) “Enhancing Cyberbullying Detection on Social Media Using Transformer
Models,” TIMES-iCON, IEEE, Bangkok, Thailand, pp. 1–5. https://doi.org/10.1109/TIMES-iCON61890.2024.10630719
[10] Wiranto, Harjito B (2023) “Enhancing machine learning performance in cyberbullying detection through hyperparameter optimization,” Proceedings of the IEEE International Conference on Technology, Engineering, and Computing Applications (ICTECA).
https://doi.org/10.1109/ICTECA60133.2023.10490843
[11] Lakshminadh K., Sravanthi V., Koushik K., and Bhaskar C. S. (2023) “Enhancing Profanity Detection in Textual Data Using BiLSTM,” ICSSAS, IEEE,
Erode, India, pp. 1–6. https://doi.org/10.1109/ICSSAS57918.2023.10331695
[12] Ea P., Xiang J., Salem O., and Mehaoua A. (2023) “Evaluating Cyberbullying Detection Algorithm Performance in Text and Image Analysis,” MLCR, IEEE, Nanjing, China, pp. 30–35. https://doi.org/10.1109/MLCR61158.2023.00015
[13] Gongane V. U., Munot M. V., and Anuse A. (2023) “Explainable AI for Reliable Detection of Cyberbullying,” PuneCon, IEEE, Pune, India, pp. 1–6.
https://doi.org/10.1109/PuneCon58714. 2023.10450132
[14] Sarkale D. G., Gabani V. J., Zhang W., and Akilan T. (2023) “NLP-driven Content Classification Towards Fake News and Bully Detection,” AIBThings, IEEE, Mount Pleasant, MI, USA, pp. 1–5. https://doi.org/10.1109/AIBThings58340.2023.10292460
[15] Jamjoom A. A., Karamti H., Umer M., Alsubai S., Kim T.-H., and Ashraf I. (2024) “RoBERTaNET: Enhanced RoBERTa Transformer Based Model for Cyberbullying Detection With GloVe Features,” IEEE Access, vol. 12, pp. 58950–58959. https://doi.org/10.1109/ACCESS.2024.3386637
[16] Obaid M. H., Guirguis S. K., and Elkaffas S. M. (2023) “Cyberbullying Detection and Severity Determination Model,” IEEE Access, vol. 11, pp. 97391–97399. https://doi.org/10.1109/ACCESS.2023.3313113
[17] Al-Hashedi M., Soon L.-K., Goh H.-N., Lim A. H. L., and Siew E.-G. (2023) “Cyberbullying Detection Based on Emotion,” IEEE Access,
vol. 11, pp. 53907–53918. https://doi.org/10.1109/ACCESS.2023.3280556
[18] Teng T. H. and Varathan K. D. (2023) “Cyberbullying Detection in Social Networks: ML vs Transfer Learning, IEEE Access,” vol. 11, pp. 55533–55560. https://doi.org/10.1109/ACCESS.2023.3275130
[19] Elsafoury F., Katsigiannis S., Pervez Z., and Ramzan N. (2021) “When the Timeline Meets the Pipeline: A Survey on Automated Cyberbullying Detection,” IEEE Access, vol. 9, pp. 103541–103563. https://doi.org/10.1109/ACCESS.2021.3098979
[20] Mansur Z., Omar N., and Tiun S. (2023) “Twitter Hate Speech Detection: A Systematic Review,” IEEE Access, vol. 11, pp. 16226–16249. https://doi.org/10.1109/ACCESS.2023.3239375
[21] Wang J., Fu K., and Lu C.-T. (2020) “SOSNet: A Graph Convolutional Network Approach to Fine-Grained Cyberbullying Detection,” IEEE Big
Data, IEEE, Atlanta, GA, USA, pp. 1699–1708. https://doi.org/10.1109/BigData50022.2020.9378065
[22] Philipo A G, et al. (2025) “Cyberbullying detection: exploring datasets, technologies and challenges,” ACM Computing Surveys. https://doi.org/10.1145/3785654
[23] Wang J., Fu K., and Lu C.-T. (2020) “Fine-Grained Balanced Cyberbullying Dataset,” IEEE Dataport, November 13. https://doi.org/10.
21227/kn1c-zx22
[24] Farouk Z, Kamel B S, Mohamed B (2013) “Adaptive backstepping control for a class of uncertain single input single output nonlinear systems,” Proceedings of the 10th International Multi-Conference on Systems, Signals & Devices (SSD), Hammamet, Tunisia, March 18–21, 2013. https://doi.org/10.1109/SSD.2013.6564134
[25] Boulkroune A, Hamel S, Zouari F, Boukabou A, Ibeas A (2017) “Output-feedback controller based projective lag-synchronization of uncertain chaotic systems in the presence of input nonlinearities,” Mathematical Problems in Engineering, Vol. 2017, Article ID 8045803. https://doi.org/10.1155/2017/8045803
[26] Rigatos G, Abbaszadeh M, Sari B, Siano P, Cuccurullo G, Zouari F (2023) “Nonlinear optimal control for a gas compressor driven by an induction motor,” Results in Control and Optimization, Vol. 11, Article 100226. https://doi.org/10.1016/j.rico. 2023.100226
[27] Charabi I, Abidine M B, Fergani B, Oussalah M (2025) “DeepF-SVM: a new hybrid deep learning model for enhanced sensorbased human activity recognition,” Cluster Computing, Vol. 28, Article 910. https: //doi.org/10.1007/s10586-025-05636-y
[28] Nuanmeesri S (2021) “A hybrid deep learning and optimized machine learning approach for rose leaf disease classification,” Engineering, Technology & Applied Science Research, vol. 11, no. 5, pp. 7678–7683. https://doi.org/10.48084/etasr.4455
[29] Goyal A, Bengio Y (2022) “Inductive biases for deep learning of higher-level cognition,” Proceedings of the Royal Society A, Vol. 478, Article 20210068. https://doi.org/10.1098/rspa.2021.0068
[30] Kishore S, He H (2024) “Unveiling divergent inductive biases of LLMs on temporal data,” Proceedings of the 2024 Conference of the North American Chapter of the Association for Computational Linguistics: Human Language Technologies (Volume 2: Short Papers), pp. 220–228, Mexico City, Mexico. https://doi.org/10.18653/v1/2024.naacl-short.20
DOI:
https://doi.org/10.31449/inf.v50i15.9750Downloads
Published
Issue
Section
License
Authors retain copyright in their work. By submitting to and publishing with Informatica, authors grant the publisher (Slovene Society Informatika) the non-exclusive right to publish, reproduce, and distribute the article and to identify itself as the original publisher.
All articles are published under the Creative Commons Attribution license CC BY 3.0. Under this license, others may share and adapt the work for any purpose, provided appropriate credit is given and changes (if any) are indicated.
Authors may deposit and share the submitted version, accepted manuscript, and published version, provided the original publication in Informatica is properly cited.
