A Truncated SVD Framework for Online Hate Speech Detection on the ETHOS Dataset

Abstract

Hate content on social media is currently one of the most significant risks, where the victim is either a single individual or a group of people. In the current scenario, online web platforms are one of the most prominent ways to contribute to an individual's opinions and thoughts. Free sharing of ideas on an event or situation also bulks on the web. Information sharing is sometimes a bane for society if primarily used platforms are utilized with some lousy intention to spread hatred for intentionally creating chaos/ confusion among the public. Users take this as an opportunity to spread hate to get some monetary benefits, the detection of which is of paramount importance. This article utilizes the concept of truncated singular value decomposition (SVD) for detecting hate content on the ETHOS (Binary-Label) dataset. Compared with the baseline results, our framework has performed better in various machine learning algorithms like SVM, Logistic Regression, XGBoost, and Random Forest.