XSentiment-HS: Hierarchical CNN-BiGRU-SVM with Explainable for Indonesian Multi-Level Hate Speech Detection
DOI:
https://doi.org/10.65897/jais.v2.i1.81Keywords:
CNN-BiGRU, explainable AI, hate speech detection , hierarchical classificationAbstract
Hate speech detection in social media requires complex text interpretation due to its spontaneous and ambiguous nature, particularly in the Indonesian language, which is rich in slang and cultural nuances. A significant challenge in current research is the predominance of binary classification models that fail to identify the severity levels of detected hate speech. To address this limitation, this study proposes XSentiment-HS, a two-stage hierarchical deep learning model for multi-level hate speech detection. The architecture integrates Convolutional Neural Networks (CNN) for local feature extraction and Bidirectional Gated Recurrent Units (BiGRU) to capture long-range contextual dependencies. The model is further enhanced with a Multi-Head Attention mechanism and utilizes a Support Vector Machine (SVM) as the final classifier. Through this integration, XSentiment-HS is expected to address feature extraction and polysemy challenges more effectively than conventional methods.
References
[1] M. O. Ibrohim and I. Budi, "Multi-label hate speech and abusive language detection in Indonesian Twitter," in Proc. 3rd Workshop on Abusive Language Online (ALW3), Florence, Italy, 2019, pp. 46–57.
[2] F. Poletto, V. Basile, M. Sanguinetti, C. Bosco, and V. Patti, "Resources and benchmark corpora for hate speech detection: A systematic review," Lang. Resour. Eval., vol. 55, no. 2, pp. 477–523, 2021.
[3] M. Selvi and S. V. N. Santhosh Kumar, "An intelligent bi-directional gate recurrent neural network based hybrid deep learning model for text-based sentiment analysis," Int. J. Comput. Intell. Syst., vol. 19, p. 114, 2026, doi: 10.1007/s44196-026-01244-9.
[4] T. Mikolov, I. Sutskever, K. Chen, G. Corrado, and J. Dean, "Distributed representations of words and phrases and their compositionality," Adv. Neural Inf. Process. Syst., vol. 26, pp. 3111–3119, 2013.
[5] N. V. Chawla, K. W. Bowyer, L. O. Hall, and W. P. Kegelmeyer, "SMOTE: Synthetic minority over-sampling technique," J. Artif. Intell. Res., vol. 16, pp. 321–357, 2002.
[6] S. M. Lundberg et al., "From local explanations to global understanding with explainable AI for trees," Nat. Mach. Intell., vol. 2, no. 1, pp. 56–67, 2020.
[7] M. T. Ribeiro, S. Singh, and C. Guestrin, "Beyond accuracy: Behavioral testing of NLP models with CheckList," in Proc. ACL 2020, pp. 4902–4912.
[8] S. Jain and B. C. Wallace, "Attention is not explanation," in Proc. NAACL-HLT 2019, pp. 3543–3556.
[9] M. S. Başarslan and F. Kayaalp, "MBi-GRUMCONV: A novel multi Bi-GRU and multi CNN-based deep learning model for social media sentiment analysis," J. Cloud Comput., vol. 12, no. 1, p. 5, 2023.
[10] C. N. Dang, M. N. Moreno-García, and F. De la Prieta, "Hybrid deep learning models for sentiment analysis," Complexity, vol. 2021, p. 9986920, 2021.
[11] B. Vidgen and L. Derczynski, "Directions in abusive language training data: Garbage in, garbage out," PLOS ONE, vol. 15, no. 12, p. e0243300, 2020.
[12] F. Koto, A. Rahimi, J. H. Lau, and T. Baldwin, "IndoLEM and IndoBERT: A benchmark dataset and pre-trained language model for Indonesian NLP," in Proc. COLING 2020, pp. 757–770.
[13] S. Wiegreffe and Y. Pinter, "Attention is not not explanation," in Proc. EMNLP-IJCNLP 2019, pp. 11–20.
[14] G. Rizos, K. Hemker, and B. Schuller, "Augment to prevent: Short-text data augmentation in deep learning for hate speech classification," in Proc. CIKM 2019, pp. 2775–2779.
[15] K. L. Tan, C. P. Lee, K. M. Lim, and K. S. M. Anbananthen, "Sentiment analysis with ensemble hybrid deep learning model," IEEE Access, vol. 10, pp. 103694–103704, 2022.
[16] A. He and M. Abisado, "Text sentiment analysis of short comments based on BERT-CNN-BiLSTM-Att model," IEEE Access, vol. 12, pp. 45229–45237, 2024.
[17] N. A. Angraini and K. M. Lhaksmana, "Sentiment analysis about legislative elections using deep learning with LSTM and CNN models," Build. Inf. Technol. Sci. (BITS), vol. 6, no. 1, pp. 289–299, 2024.
[18] A. Alsayat, "Improving sentiment analysis for social media applications using an ensemble deep learning language model," Arab. J. Sci. Eng., vol. 47, no. 2, pp. 2499–2511, 2022.
[19] S. Ahmad, S. M. Saqib, and A. H. Syed, "CNN and LSTM based hybrid deep learning model for sentiment analysis on Arabic text reviews," Mehran Univ. Res. J. Eng. Technol., vol. 43, no. 2, pp. 183–194, 2024.
[20] A. Aslan, S. Kızıloluk, and E. Sert, "TSA-CNN-AOA: Twitter sentiment analysis using CNN optimized via arithmetic optimization algorithm," Neural Comput. Appl., vol. 35, no. 14, pp. 10311–10328, 2023.
Downloads
Published
Issue
Section
License
Copyright (c) 2026 Gina Khayatun Nufus, Rizki Dewantara, Ardi Susanto, Sokid, Lia Farhatuaini, Jaka Septiadi, Mohammad Raihan Akbar (Author)
This work is licensed under a Creative Commons Attribution-ShareAlike 4.0 International License.
This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution-ShareAlike 4.0 International License (CC BY-SA 4.0).
Under this license, users are permitted to copy, distribute, display, and perform the work, and to make derivative works, provided that appropriate credit is given to the author(s) and the source, and that any derivative works are distributed under the same license.
