Toxicity Tweet Detection and Classification Using NLP Driven Techniques
Anuj Kumar Pal, Sakshi Rai · 2023
The use of social media is increasing regularly. Unethical things like defamation, spreading hatred, pornography, etc are becoming easier for some irresponsible users because of the easy accessing and due to similarity of social media. In this view, this study has been carried out to use machine learning techniques to classify comments into their hazardous categories. In order to categorize a comment based on its toxicity, this research compares classic machine learning methods with deep learning methods including Logistic Regression, SVM, RNN, and LSTM. To compare the performance of the models, four distinct models are developed, put into use, and trained on a common dataset. These models were developed and evaluated using sizable amounts of secondary qualitative data that included several comments that were either labeled as harmful or not. Results showed that employing LSTM, a satisfactory accuracy of 90.7% and an F1- score of 0.94 were obtained.