The identification of hate speech has emerged as a crucial responsibility for upholding civil debate and guaranteeing user security on Internet platforms. This paper presents a technique that uses a dataset that has been adversely evaluated by people to identify hate speech in Wikipedia comments. Convolutional neural networks and bidirectional long short-term memory (BiLSTM) networks are the primary models utilized in this study for the categorization of hate speech. In this paper, strategies are being used to address imbalanced classes and preprocess the data to obtain pertinent features. It is shown through considerable experimentation that our suggested approach is effective in accurately detecting instances of hate speech. To demonstrate the superiority of our model in terms of performance measures like precision and recall this paper also does a comparative analysis with existing approaches. Results support the continuous efforts to create a more welcoming and courteous online environment by highlighting the significance of utilizing cutting-edge DL algorithms for identification of hate speech in online forums.