DEEP LEARNING BASED CYBERBULLYING DETECTION ON SOCIAL MEDIA PLATFORMS

Abstract

Cyberbullying has become a major problem in social media, which poses severe psychological and emotional dangers to the users, especially adolescents and young adults. The explosive development of online communication has resulted in a rise in abusive and harmful content, which makes manual monitoring inefficient and impractical. The current systems of cyberbullying detection, which are mainly founded on the traditional machine learning techniques, are highly reliant on the detecting systems based on the analysis of keywords and lack the capacity to detect the contextual meaning, sarcasm, and linguistic variations, which contributes to the decreased accuracy and reliability. Recent works emphasize that deep learning methods are more successful than traditional ones because they are capable of capturing semantic and contextual relationships in textual data. The proposed research involves a deep learning-based system to detect cyberbullying based on a Bidirectional Long Short-Term Memory (BiLSTM) model. The system uses Natural Language Processing (NLP) methods that include text cleaning, text tokenization, removal of stop-words, and text normalization to preprocess text in social media. The word embeddings are used to convert the textual data to meaningful numerical representations so that the model can learn the semantic relationships between words. The BiLSTM architecture handles textual data both forward and backward, enabling it to effectively deal with contextual information and sequential dependencies. The proposed model is a multiclassification to discover the various categories of cyberbullying, in this case, age-based, gender-based, religious, and ethnic abuse. Experimental data has shown that the BiLSTM model can achieve the highest accuracy, precision, recall and F1-score than the traditional machine learning models. The system is scalable and can be deployed in real-time, which can help create safer online spaces as it can help detect and prevent cyberbullying content on the internet effectively.