Paper Title
Online Safety Enhancement: A Real-Time Hate Speech Detection System Using BERT, Ensemble Learning, and Traditional Algorithms in Stream Lit

Abstract
This project specifically aims at identifying hate speech in social media content using BERT and classical machine learning methods, Naive Bayes, Logistic Regression, SVM and Random Forest. Textual information is processed through methods of tokenization, removal of stop words, lemmatization and then determined if it is hate speech or not. The BERT Classifier is trained and tested using a torch with metrics of accuracy and confusion matrices for enhanced measurement. Several model types are compared with each other and combined in order to increase the overall prediction accuracy. Word clouds and accuracy plots are additional features to give more information on how the model is performing. This Stream lit app allows the user to engage with the system for the real-time detection of hate speech. The project is designed to develop a dependable and efficient system for monitoring and moderating what is considered toxic especially in online communities with consideration for context. Keywords - Hate Speech Detection BERT Naive Bayes Logistic Regression SVM Random Forest Streamlit Machine Learning NLP Torch Real Time Detection.