Mitigation of Online Public Shaming Using Machine Learning Framework

Abstract

In the digital world, currently some of the most far-reaching sites are social media sites on the internet. Billions
of users are associated with social network sites. User interactions with these social sites, like twitter has an enormous and
occasionally undesirable impact implications for daily life. Large amount of unwanted and unrelated information gets spread
across the world using online social networking sites. Twitter is one of the most extensive platforms and it is the most popular
micro blogging services to connect people with the same interests. Due to the popularity of twitter, it becomes a main target
for shaming activities. Nowadays, Twitter is a rich source of human generated information which includes potential
customers which allows direct two-way communication with customers. It is noticed that most of the participating users post
comments in a particular occurrence are likely to embarrass the victim. Interestingly, it is also the case that shaming whose
follower counts increase at higher speed than that of the nonshaming in Twitter. The proposed system allows users to find
disrespectful words and their overall polarity in percentage is calculated using machine learning algorithm. Shaming tweets
are grouped into nine types: abusive, comparison, religious, passing judgment, jokes on personal issues, vulgar, spam, nonspam and whataboutery by choosing appropriate features and designing a set of classifiers to detect it.