Offensive Tweets Detection in Hinglish Using HingBERT
摘要
Social media is the perfect platform for people to interact, share news, and express their own thoughts. Unfortunately, there is a lot of negativity in it, including hate and abusive speech. People in multilingual nations like India frequently blend two or more languages to create hybrids. Hinglish is an example of a hybrid language that combines Hindi and English. This research concentrates on identifying offensive Hinglish-language tweets. Fine-tuning on HingBERT, HingMBERT, and HingRoBERTa has been done to improve upon the existing results. With a 0.95 F1-score, HingRoBERTa achieved benchmark results, while HingMBERT and HingBERT scored 0.93 and 0.93, respectively.