Emotion Recognition In Conversation

Emotion Recognition in Conversation (ERC) is a subtask of Natural Language Processing that aims to identify the emotional labels of each utterance from predefined emotion categories by analyzing the conversational transcript text and its speaker information. Specifically, given a sequence of N utterances [(u1, p1), (u2, p2), ..., (uN, pN)], where each utterance ui consists of T words spoken by the speaker pi, the goal of the ERC task is to predict the corresponding emotional label ei for each utterance ui. This task is of significant value for enhancing human-computer interaction experiences, sentiment analysis, social robots, and other applications.