An empirical study of Multimodal Entity-Based Sentiment Analysis with ChatGPT: Improving in-context learning via entity-aware contrastive learning

Li Yang, Zengzhi Wang, Ziyan Li, Jin-Cheon Na, Jianfei Yu

Abstract: Highlights•The first work to explore in-context learning (ICL) for the MEBSA task;•Develop an ICL approach with entity-aware contrastive learning to retrieve similar samples;•Devise a similarity scoring function to construct positive and negative instances for model training.