Understanding Cyberbullying Patterns Utilizing Word Clouds

Social media provides a convenient way to communicate among people by sending and receiving content. However, children and teenagers may suffer from emotional, mental, and behavioral problems due to its negative influences, such as unpleasant, damaging, deceiving, or malicious content towards an individual, referred to as cyberbullying. Thus, detecting suspicious conversations in cyberspace is essential for users to increase their awareness. This research aims to identify important features from cyberspace texts to generate a predictive model. As an initial step of the study, we focus on understanding the patterns between cyberbullying and non-cyberbullying utilizing word clouds. In detail, after pre-processing steps such as text cleaning, stopword removal, and lemmatization, we compared two measurements, term frequency-inverse document frequency (TF-IDF) and Bag-Of-Words (BoW), to determine the importance of words in identifying cyberbullying by selecting the most frequently used words. Word clouds are then created to present the selected words visually. The determined words could be used as essential indicators to determine cyberbullying conversation. We also compared emotion-related words between cyberbullying and non-cyberbullying texts. This study can provide a deeper understanding of language patterns related to cyberbullying conversations.