Hate Speech Detection in Telugu Language Using Transformer Models and Machine Learning

Hate speech in the social media raises a fundamental concern of social integration and quality individual life. This paper aims at presenting the key importance of hate speech detection in the Telugu language which is a Dravidian language and officially used in the Indian states. Through the use of various transformer models such as BERT, DistilBERT, indicBERT, LaBSE, and MuRIL, the proposed solution offers acceptable accuracy, especially with the MuRIL imprinting an Fl-score of 0.79. The combination of ensemble embedding with machine learning classifiers escalates the performance and exhibits a better position to advance content moderation for only languages in the mainstream multilingual online platforms. This research not only addresses a relevant problem but also helps demonstrate the possibilities of using NLP developments to deal with online hate speech, thus creating less dangerous digital landscapes. The results also underscore the need to maintain the development process and update the further strategies to identify the hate speech persistently due to new tendencies in the interaction within social media platforms.