Detecting Hateful and Offensive Language

Sparked by the alarming situation on social media platforms, where there is a dramatic increase in inflammatory language, companies have already implemented algorithms to regulate or even remove extreme posts. On the other hand, freedom of opinion and expression is a cornerstone of many societies, raising concerns that attempts to curb inappropriate language could also lead to the restraint of free speech. The current chapter aims to identify hateful and offensive language in tweets. Without delving into the particulars of this debate, we will address a few technical challenges and provide possible solutions in this setting. During this process, we also introduce many new concepts and techniques for machine learning.

A central theme of this chapter concerns the reuse and tuning of third-party models to minimize the effort of a new deployment. Using an open source dataset with hateful and offensive tweets, we will examine the steps to build...