Hate Speech Detection and Marker Identification Approach

Hate speech detection has been viewed as the need of the hour with the sharp rise of social media, where destructive content can spread both quickly and widely. Traditional detection methods fall behind in handling diverse and context-dependent expressions of hate speech. The proposed framework for hate speech detection involves attention-based deep learning model and identification of hate speech markers. We experiment with benchmark datasets and compare with SOTA approaches, and our model is better than SOTA in terms of accuracy, precision, recall, and F1-score. We also test the model with audio to check if it is able to identify the hate speech markers.