Identification of Offensive Language in Social Media Using Prompt Learning

Offensive language refers to the use of language in a manner that may offend or harm others who are within earshot or view in a public place. Given the importance of identifying such language in social media for promoting emotional well-being, we propose a prompt learning method and compare its performance with fine-tuning on two widely used datasets, HatEval and OffensEval. Experimental results demonstrate that prompt learning can achieve a performance improvement over fine-tuning in a fully supervised setting.