A PRACTICAL APPROACH TO CATEGORIZE HATE SPEECH IN ONLINE SOCIAL NETWORKS
The main objective of this paper is to analyze the performance in terms of accuracy for various
Machine Learning (ML) models such as Support Vector Machines (SVM), Logistic regression (LR),
and ensemble classifier to compartmentalize the hate speech on Online Social Networks. It deals with
a process centered on the automatically obtaining patterns and unigrams through training dataset.
Those unigrams and patterns can be used afterward as a preparation (training) for algorithms in
machine learning, among others. Sentiment Analysis is used to detect the polarity of the tweets if it is
clean, hatred, or offensive by using sentiment analyzer. The system is implemented on a range of
24783 tweets. The results prove that the deployed model attained an accuracy rate of 88%, 72%, and
66% using SVM, LR, and Ensemble classifier respectively. The binary and ternary classification was
implemented on tweets to categorize the hate, offensive, and neutral speeches.