Hate Speech Detection Using Deep Learning Models

This research addresses the critical problem of hate speech proliferation on social media platforms, which can lead to real-world harm, including violence and marginalization of vulnerable communities. The goal of this study is to evaluate the effectiveness of advanced machine learning models-BERT, RoBERTa, and LSTM-in detecting hate speech from social media data. Two datasets with different characteristics were used to build the model. They were put through few preprocessing steps which includes tokenization, lemmatization, and TF-IDF feature extraction. Transformer based models outperform the traditional RNN based model - LSTM with RoBERTa achieving the highest accuracy of 93%. Our results demonstrate that these advanced models are capable of recognizing hate speech with much certainty, thus offering hope for automated content moderation.