Scalable Deep Learning Frameworks for Cyberbullying Detection in Social Media

The increased spread of social media networks has compounded the problem of monitoring and controlling cyberbullying, hate speech, and offensive language on a largescale basis. Conventional detection systems are not able to handle the large quantity of unstructured text and multimedia data being created on a daily basis, in addition to guaranteeing privacy and computational performance. The study resolves such difficulties by combining the deep learning Darwens such as convolutional neural networks (CNNs), gated recurrent units (GRUs), and transformer-based frameworks with the big data model that can work with large-scale data like Apache Spark and MapReduce. The suggested method builds upon the principles of distributed in-memory processing and RDD-based anonymization to provide the ability to analyze mass data in a privacy-preserving and high-throughput fashion. Empirical testing indicates that the scalability of the anonymization pipelines and deep learning classifiers together yield high gains in accuracy of detection as well as system performance, especially in the multi-class cyberbullying contexts. Moreover, the framework facilitates cultural flexibility through the integration of sentiment analysis and bias detection modules so that it is robust in various linguistic and social contexts. The findings demonstrate the possibility of implementing end-to-end scalable systems that use AI as a tool to detect cyberbullying in real-time as a technologically sound solution to the challenge of delivering secure, efficient, and socially responsible online communities.

Ҳали таржима қилинмаган