Empathetic Dialogues context labels
Empathetic Dialogues is a public corpus from Facebook AI of conversational utterances paired with one of 32 emotion-context labels, designed to train empathetic chat agents. The texts in this corpus frequently explore themes of personal achievements , emotional support , and interpersonal relationships , reflecting a range of experiences and sentiments. Common topics include academic milestones , such as graduation and job acquisition, alongside expressions of concern for loved ones facing challenges, like health issues or accidents. Complaints about trust and reliability in relationships also emerge, as do moments of joy and celebration related to personal events or milestones. Additionally, the texts convey a sense of community through shared experiences and advice, highlighting the importance of empathy and understanding in everyday interactions. [Summary on 50 random texts by ChatGPT 4o Mini].
Reported holdout systems from the verified card
Multiclass classification uses Macro-F1 as the task-primary metric. Secondary columns keep the companion metrics visible so binary, ordinal, regression, and multiclass cards are not compared through one flattened score.