A Review of Deep Learning and NLP Based Techniques Used for Detection of Hate Content Against LGBTQ+ on Social-Media

ABSTRACT

This paper explores ten research studies focused on detecting hate content against LGBTQ+ individuals on social media using deep learning and natural language processing (NLP). The review examines methodologies, datasets, and outcomes, identifying key trends and limitations. Techniques such as transformer-based models (e.g., BERT, RoBERTa) and ensemble methods dominate the landscape, demonstrating superior performance in multilingual and code-mixed language settings. However, challenges such as detecting implicit hate speech, handling linguistic diversity, and addressing class imbalance persist. To address these issues, this paper proposes a hybrid model combining BERT and GRU-RNN. This approach leverages BERT’s contextual embeddings for nuanced understanding and GRU’s efficiency in sequential data processing, enhancing detection accuracy and scalability. The model’s adaptability to multilingual datasets and ability to handle long-language sequences make it a robust solution for diverse social media platforms. Future research should focus on domain-specific datasets, explainable AI integration, and real-time implementation to combat hate speech effectively. By bridging existing gaps, this review aims to advance the discourse on hate speech detection and foster safer, more inclusive online environments for LGBTQ+ communities.

FULL TEXT