An Exploration of Machine Learning and Deep Learning Techniques for Offensive Text Detection in Social Media-A Systematic Review

The increasing popularity of usage of social media platforms such as Facebook, Twitter, and What's App has also given a potential to spread hatred or to cause harassment or inconvenience by using offensive and abusive texts on these platforms. It has been identified that offensive language is a significant problem for the safety of both social platforms and their users. The circulation of offensive or abusive language to the online community undermines its reputation, scares away users and also directly affects their mental growth. Offensive or abusive text just not only affects users but also affects stakeholders such as governments, autonomous organizations, and social media platforms. Every day such stakeholders have to spend long hours to remove such content manually from these platforms. So, there arises the need to detect offensive and abusive text in user's posts, messages, comments, blogs, etc., automatically. To address this issue, detection of offensive/abusive text in user's message, posts, comments, blogs, etc., has become a crucial task in recent times. There are various machine-learning and deep learning approaches existing in literature to identify such abusive texts. We have followed a systematic review process, in which we aim to explore the various machine learning or deep learning approaches adopted by various researchers to detect and the offensive/abusive speech in user's textual posts, messages, comments, blogs, etc. This systematic review will help to strengthen the design and implementation of a new and efficient approach for automatic detection and removal of abusive or offensive text in user's message or post. This deep exploration of the existing techniques will further have strong benefit to people, society, government, and social platforms in order to avoid spreading of hatefulness, harassment through social media.