Evaluating Machine Learning Models for Hate Speech Detection in ODIA Language

Low-resource language research can strengthen local communities by providing them with a platform on the international scene. ODIA can be considered as one of the low-resource languages as it is confined only to the region of Odisha. This study delves into the identification of hate speech for the ODIA language on social media platforms through the utilization of machine learning algorithms. In this study we deploy and evaluate different machine learning modals for hate speech detection in ODIA language. Our study employs TF-IDF for Feature extraction and the different well known classification alogrithm i.e SVM, Logistic Regression, Random Forest, Gradient Boosting and Adaboost. The results demonstrate that TF-IDF features, particularly when combined with SVM, exhibit a high level of accuracy (0.838) in detecting hate speech within ODIA datasets. The findings of the study addresses the automatic identification of hate speech in ODIA language can be extracted to other low-resource languages.