Enhancing Hate Speech Detection on Social Media through Sentiment Analysis and Transformer-Based Deep Learning Techniques
Muhammad Erlangga Putra Suryono, Arif Djunaidy · 2024
The widespread use of social media platforms like Facebook, Twitter, Instagram, and YouTube has fundamentally changed the way we communicate, allowing for the rapid exchange of information and fostering a sense of connectivity like never before. However, the lack of oversight on these platforms has also led to significant issues, such as the spread of misinformation, fake news, and hate speech. Our research aims to tackle these problems by integrating sentiment analysis with hate speech detection to enhance the performance of models analyzing social media data. We delve into recent studies and technological advancements to shed light on the benefits and limitations of sentiment analysis in this area. We conducted experiments using various deep learning models, including RNN, CNN, and LSTM, to determine the impact of sentiment analysis on detecting hate speech. Our findings reveal that incorporating sentiment analysis improves the accuracy of these models. Notably, the LSTM model achieved the highest accuracy at 69.77%when combined with sentiment analysis, a significant improvement from its baseline accuracy of 67.52%. Moreover, we found that using BERT word embeddings alongside sentiment analysis further boosts performance, underscoring the importance of advanced NLP techniques in creating effective hate speech detection systems. This research highlights the critical role that combining sentiment analysis with deep learning can play in better identifying and mitigating hate speech on social media, marking an important step forward in this field.