Cyberbullying Detection Using Continuous Based Bag of Words with Machine Learning by Text Classification

Abstract

The breakneck advancement in internet for Social Media (SM) have generated enormous text data that became a challenging as well as valuable task in identifying an adequate measure to analyze text data using machine. Natural Language Processing (NLP) technique is one of the text classification methods that applicable for several applications sectors such as e-commerce and customer service. Bulling over SM for individuals have resulted with calumny, chastise and threatening. This kind of cyberbullying generates increase in serious mental health issues particularly for young generation which resulted to lessened self-esteem as well as increase of suicidal reflection. A generation of young adults will be affected by mental health and self-esteem problems, if action is not taken to stop cyberbullying. However, cyberbullying has become the ultimate challenge for Artificial Intelligence (AI) studies as well as more beneficial in the real-life applications. Therefore, initial step of the machine is for understanding the text by text representation whereas the most preferable method is Bag-of-Words (BoW). This paper has proposes a Continuous Based BoW (CBBoW) method assist to perform better significance for minimizing the training time requirement and even accomplish the training accuracy rate. The results determine that suggested method accomplishes performance with best accuracy in detection of cyberbullying words. The suggested techniques are tested using conventional BoW and Word2Vec approaches on open-source datasets with predetermined data partitions provided accessible through an open digital repository to promote replication.