A STUDY ON THE PERFORMANCE OF NLP-BASED MODELS IN ABUSIVE CONTENT CLASSIFICATION

This paper explores the efficiency of various machine learning models for abuse detection in text, comparing traditional models (Logistic Regression, Random Forest, Decision Trees) with advanced deep learning techniques (RNNs, LSTMs, Bi-LSTMs, CNNs) and pretrained transformers (BERT, RoBERTa, DistilBERT, XLNet). The study also investigates hybrid models that combine the strengths of these individual approaches to improve accuracy. By evaluating the performance of these models on abuse detection tasks, the research aims to identify the most effective methods for automatically detecting abusive language in online content, contributing to more efficient content moderation systems. Keywords: Abuse Detection Models, Offensive Language Classification, Text Classification, NLP, Abuse Detection, Deep Learning for Toxicity Detection