In recent years, many people on the internet write and post abusive language on online social media platforms such as Twitter, Facebook, etc. Detection of hate speech is very difficult to solve manually, especially in social media. Thus, we need to be automatic detection of hate speech in social media. We have used a benchmark dataset of approximately 25 thousand annotated tweets and proposed a model based on deep learning methods. We also compare the performance of our deep learning methods to the traditional machine learning classifier in terms of F1 Score and Accuracy. The results obtained through the proposed method is very promising.
The uses of offensive languages on social media's like Instagram, Facebook, Twitter, etc. are increased tremendously. Nowadays, huge growth in the number of social media users are seen. Peoples are knowingly or unknowingly flooded the social media platforms with offensive posts. This become very challenging task to detect offensive posts. Manually solving this problem is not feasible, therefore automatic detection of offensive language is needed. Our objective of the work is to detect offensive language with the highest accuracy. The Davidson dataset is used for experiments with annotated tweets and implemented based on LSTM and BERT methods. The proposed deep learning methods is compared with other known machine learning classifier. Overall, result analysis is show that proposed deep learning method is outperforming others.