Automatic Hate Speech Detection Based on BERT and CNN

With the rise of social media, billions of users post and browse various topics and content daily. Text data contains rich information, reflecting users' opinions and attitudes towards products, services, events, and more. Detecting hate speech and offensive language in text helps platform administrators, governments, and organizations quickly identify and address inappropriate content, promoting a civilized and harmonious online environment. This paper proposes an automatic method for detecting hate speech and offensive language based on BERT and CNN. First, the pre-trained BERT model converts the text into vector representations. Then, the CNN extracts text features, followed by classification using a softmax layer. Experiments on a public hate speech and offensive language detection dataset demonstrate the effectiveness and superiority of the proposed method compared to other commonly used approaches. This method not only accurately detects hate speech and offensive language but also provides valuable insights for platform administrators, aiding them in better handling such issues.