TY - GEN
T1 - A Comprehensive Study of Text Classification Models for Hate Speech Detection in Resource-Limited Environments
AU - Reddy, Sai Mahith
AU - Loiya, Daksh
AU - Kulal, Srinivas S.
AU - Reddy, G. Pradeep
AU - Raghavendra, S.
N1 - Publisher Copyright:
© 2025 IEEE.
PY - 2025
Y1 - 2025
N2 - Social media platforms generate vast amounts of user content requiring automated sentiment analysis and offensive language detection. This study evaluates six classification algorithms-GRU, LSTM, CNN, Transformer, TCN, and XGBoost-using a Twitter dataset with binary offensive/non-offensive labels. Two preprocessing approaches (raw text versus noun-filtered text) were combined with TF-IDF and BERT feature extraction methods. Results demonstrate that raw text preprocessing consistently outperformed noun filtering across all models. TF-IDF vectorization achieved superior accuracy and efficiency compared to BERT embeddings. The optimal configuration combined GRU architecture with TF-IDF features, achieving 79.2% accuracy, 88.9% AUC, and 78.9% F1-score. These findings establish the GRU-TF-IDF approach as an effective and computationally efficient solution for practical sentiment analysis applications.
AB - Social media platforms generate vast amounts of user content requiring automated sentiment analysis and offensive language detection. This study evaluates six classification algorithms-GRU, LSTM, CNN, Transformer, TCN, and XGBoost-using a Twitter dataset with binary offensive/non-offensive labels. Two preprocessing approaches (raw text versus noun-filtered text) were combined with TF-IDF and BERT feature extraction methods. Results demonstrate that raw text preprocessing consistently outperformed noun filtering across all models. TF-IDF vectorization achieved superior accuracy and efficiency compared to BERT embeddings. The optimal configuration combined GRU architecture with TF-IDF features, achieving 79.2% accuracy, 88.9% AUC, and 78.9% F1-score. These findings establish the GRU-TF-IDF approach as an effective and computationally efficient solution for practical sentiment analysis applications.
UR - https://www.scopus.com/pages/publications/105030100654
UR - https://www.scopus.com/pages/publications/105030100654#tab=citedBy
U2 - 10.1109/DISCOVER66922.2025.11258993
DO - 10.1109/DISCOVER66922.2025.11258993
M3 - Conference contribution
AN - SCOPUS:105030100654
T3 - 2025 IEEE International Conference on Distributed Computing, VLSI, Electrical Circuits and Robotics, DISCOVER 2025 - Proceedings
SP - 152
EP - 157
BT - 2025 IEEE International Conference on Distributed Computing, VLSI, Electrical Circuits and Robotics, DISCOVER 2025 - Proceedings
PB - Institute of Electrical and Electronics Engineers Inc.
T2 - 9th IEEE International Conference on Distributed Computing, VLSI, Electrical Circuits and Robotics, DISCOVER 2025
Y2 - 17 October 2025 through 18 October 2025
ER -