Discourse-Aware Causal Emotion Entailment

Causal Emotion Entailment (CEE) aims to identify the corresponding causal utterances for a target emotional utterance in conversations. Most previous research has focused on the use of sequential encoding to model conversational contexts, without fully considering the interaction effects between different utterances. In this paper, we explore the significance of discourse parsing in addressing these interactions, and propose a new model called discourse-aware model (DAM) to tackle the CEE task. Concretely, we use a multi-task learning framework to jointly model CEE and discourse parsing to fuse rich discourse information. In addition, we use a graph neural network to further enhance our CEE model by explicitly encoding discourse and other discourse-related structure features. The results on the benchmark corpus show that the DAM outperforms the state-of-the-art systems in the literature. This suggests that the discourse structure may contain a potential link between emotional utterances and their corresponding cause expressions. We will release the codes of this paper to facilitate future research ( https://github.com/Sakurakdx/DAM ).