LECM: A model leveraging emotion cause to improve real-time emotion recognition in conversations

Wenxuan Lu, Zhongtian Hu, Jiashi Lin, Lifang Wang

Abstract: Highlights•Emotional cause information helps to improve the performance of real-time emotion recognition in conversations.•Proposed a multi-task learning model, LECM, which effectively leverages emotional cause information in conversation history.•The task of causal emotion entailment is used as an auxiliary task to identify the causes that trigger the target utterance emotion.•Designed a cause-oriented tendency perception network to guide the model to focus on important information in the conversation history.