Trusting CHATGPT: how minor tweaks in the prompts lead to major differences in sentiment classification

Open in new window