Detecting Cyberbullying and Hate Speech in Regional Languages Using Hybrid Deep Learning and NLP Models
Ganesh C.1, Kumarganesh S.2, Elayaraja P.3, Thiyaneswaran B.4
1Department of ECE, Knowledge Institute of Technology, Salem, Tamil Nadu, India
2Department of ECE, Sona College of Technology, Salem, Tamil Nadu, India
3Department of CCE, Sri Eshwar College of Engineering, Coimbatore, Tamil Nadu, India
4Department of ECE, Kongunadu College of Engineering and Technology, Trichy, Tamil Nadu,
India
Emails: csganesh86@gmail.com; sunmun@gmail.com; saikgss@gmail.com; thiyanesb@yahoo.co.in
Abstract
The rise of social media platforms has led to an increase in cyberbullying and hate speech, which can have severe consequences on individuals and communities. The detection of harmful content, especially in regional languages, remains a significant challenge due to the linguistic diversity, informal expressions, and limited datasets available for training machine learning models. This paper proposes a hybrid deep learning and natural language processing (NLP) model for the detection of cyberbullying and hate speech in regional languages. The model combines convolutional neural networks (CNNs) and recurrent neural networks (RNNs) with advanced NLP techniques such as sentiment analysis and context-aware feature extraction. Preliminary experiments show that the proposed model achieves an accuracy of 86.7% for hate speech detection and 82.3% for cyberbullying detection in regional language datasets. Furthermore, the hybrid model outperforms traditional machine learning techniques by 15% in terms of precision and recall. This approach demonstrates the potential of combining deep learning and NLP to address the challenges of detecting harmful content in diverse linguistic environments.
Keywords: Cyberbullying Detection; Hate Speech Detection; Regional Languages; Deep Learning; Hybrid Models; Natural Language Processing (NLP); Convolutional Neural Networks (CNN); Recurrent Neural Networks (RNN); Sentiment Analysis; Data Augmentation