Technical Challenges to Automated Detection of Toxic Language

These days, more people are using online platforms as communication tools. Despite the obvious benefits of expanded material sharing, the past ten years have witnessed a disturbing surge in toxic communication, such as cyberbullying and harassment. Due to its complexity and context-dependence as well as the fact that current techniques of detection mainly rely on very unpleasant language like slurs and profanity, online toxicity can be challenging to detect. The importance of the issue has sped up development in the area of automatically detecting abusive language in online content that is posted on social media. Since exposure to online toxicity can have significant societal repercussions, reliable models and algorithms are required for recognizing and evaluating such communication throughout the wide and expanding realm of social media. We comprehensively characterize the conceptual traits of conflictual online languages in this paper and also discuss the technical challenges in automatic detection of toxicity.