Cyberbullying Comments Detection and Deletion from Social Media: A Review

Abstract

Nowadays cyberbullying is increasing day by day. All the social media platforms can detect the cyberbullying comments and the comments can be hided by the channel owner. There is no system for the automatic deletion of such comments. So this review is conducted to find the best accurate algorithm for the detection of cyberbullying comments.Cyberbullying has become a giant social hassle in the virtual age, affecting people, particularly younger humans, on many online sites. Pervasive social networks and the ability to talk anonymously often make it hard to come across and deal with awful behavior. This observe investigates how machine learning can be used to stumble on cyberbullying in social media posts, and aims to create a pressure that could hit upon and reduce these negative outcomes. We first explain the which means and sorts of cyberbullying, discuss its mental impact on victims, and deal with the issue of identifying cyberbullying because of the huge amount of on line datasets. Traditional discovery strategies, such as publications or content material filters, often omit the nuances of on line language, indicating the want for solutions. Our device getting to know goal leverages machine learning (NLP) to investigate social context the usage of bullying and non-bullying labels to train numerous schooling fashions. We use metrics together with precision, consider, F1 score, and accuracy to evaluate techniques such as aid vector machines, random forests, and deep getting to know models (inclusive of LSTM and Transformer architectures). We improve the performance of our model in information content and sentiment the use of content extraction strategies (Bag of Words, TF-IDF, and phrase embeddings (Word2Vec, GloVe)). The outcomes show that deep mastering fashions, in particular LSTM and Transformer-based models, outperform strategies in online harassment detection and allow message removal.