Hate Speech Detection Model Using Bag of Words and Naive Bayes

In this era of increasing hate and intolerance among the people, especially among those who interact with each other over the Web, there is a dire need of some technological innovation that would cater to this situation. The said hate and clash of opinions among the people often comes out in the form of hate speech in texts and in pictures. To counter this situation, we have come up with a hate speech detection model which would be able to detect and identify hateful and provocative content in a textual data, which is published on various social media websites, viz. Twitter, Facebook, and Instagram. The sole idea behind the making of this model is to be able to prevent every individual from spreading as well as witnessing hate-speech on different digital-platforms. We have developed a text classifier using basic principles of natural language processing. This has been achieved by the use of the bag of words model for feature extraction purposes, followed by various text filtering processes, and ultimately feeding this data to a naive-Bayes classifier, and hence training the same to work autonomously to classify textual data depending upon the sentiments indicated by them, i.e. whether they imply negative aspects over a certain matter/topic or positive. As a result of this experiment, we were able to successfully classify all the data taken by us with a cumulative accuracy of 99.7% upon the test data set.