Challenges and innovations in machine learning-driven hate speech detection on X (formerly Twitter): A methodological and taxonomic review

Abstract

Detecting hate talk on public news manifestos, particularly X (formerly Twitter), has enhanced a bigger focus on document excavating and the study of natural language processing (NLP). This systematic literature review (SLR) offers an itemized test of current approaches, machine intelligence forms, challenges, and time engaged in hate talk discovery. The review covers 91 research documents written from 2016 to 2022, preferred through an all-encompassing election process. It evaluates a range of directed and alone machine intelligence methods, to a degree Support Vector Machines (SVM), Random Forests, Neural Networks, and fresher turbine-located models like BERT and GPT. The study also addresses detracting issues like dataset bias, model generalizability, and exposure to opposing attacks. Additionally, it labels research breaches and implies future guidance for reinforcing hate talk discovery arrangements. The understandings from this review aim to support the invention of more exact and flexible models for detecting and fighting connected to the internet hate talk