Classification of cyberbullying in social media using Natural Language Programming method

Abstract The evolution of internet describes the usage of Social Media (SM) which expands drastically. This involvement continues to increase in context with the current global epidemic since people frequently use SM platforms to vent their feelings. Similarly, the widespread adoption of SM sites like Facebook and Twitter by many organizations has raised the volume of essential individual input regarding the state of affairs, events, goods, and services. One of the major drawbacks that widespread SM usage is cyberbullying. The frequency of cyberbullying on SM platforms has raised serious concerns for people, organizations, and society at large. To minimize negative impacts due to cyberbullying on SM, early detection is essential. As a result, Sentiment Analysis (SA) utilizing Twitter data has gained prominence. Text analysis-related Natural Language Processing (NLP) and in Artificial Intelligence (AI) technologies have gained increased attention because of increasing demand for SM analysis. Moreover, the important fields utilized to proactively extract and train high-quality characteristics from low-level text involved is Machine Learning (ML). The Convolution Neural Network (CNN) models are Deep Learning (DL) methods to train data like text, picture, and video data. Improved text classification is achieved by preparing these data types using CNN as a persuasive method. To categorize the bullying text, this study discusses the Ensemble model that integrates Modified Term Frequency and Inverse Document Frequency (MTF-IDF), and Deep Neural Network (DNN) with sophisticated feature extraction approaches. From the text data, the cyberbullying feature patterns are extracted using the techniques of feature extraction.