AI Based Identification of Inappropriate Language

In today's digital age, the proliferation of online communication has led to an increased need for automated systems to identify and combat inappropriate language, hate speech, and offensive content. This study explores the effectiveness of four distinct machine learning algorithms – Decision Trees, Random Forest, Long Short-Term Memory (LSTM), and BERT – in predicting whether a given text contains offensive language. Decision Trees and Random Forest are traditional machine learning techniques that rely on structured feature engineering, while LSTM and BERT represent cutting-edge deep learning methods capable of handling unstructured text data. By comparing these algorithms, this research aims to provide insights into their respective strengths and weaknesses in addressing the evolving challenges of online content moderation. Our findings reveal the accuracy, precision, recall, and F1-score of each model, shedding light on their ability to discern inappropriate language. Ultimately, this study contributes to the development of robust AI-based content filtering systems, assisting platforms and social networks in maintaining safe and respectful online environments by automatically identifying and flagging offensive text. Such systems have the potential to foster more inclusive and responsible digital communities, ensuring a healthier online discourse for all users.

 Keywords: Random Forest, Decision Tree. LSTM and BERT.