Leveraging Two-Stream Cause-Effect Relation for Emotion-Cause Analysis

Anonymous

Abstract:

Emotion Cause Analysis (ECA) is a task to analyze corresponding causes for certain emotions expressed in text, which heavily depends on the context as the model needs to find the deep cause-effect relations between emotions and their causes.
Previous research typically focused on extracting emotions first and then their corresponding causes, or vice versa.
However, these approaches fail to integrate these two streams of thought into a unified model, so we propose a novel two-stream reasoning model to unify them for better performance.
We leverage discourse connectives as bridges between these two streams, incorporating their discourse information to reveal cause-effect relations and enhance the reasoning ability of our model.
Further, we employ the connectives predicted by ChatGPT to help our model achieve better results, and our research demonstrates that our model achieves SOTA results in ECA and proves the superiority of our model.

Paper Type: long

Research Area: Sentiment Analysis, Stylistic Analysis, and Argument Mining

Contribution Types: Model analysis & interpretability, NLP engineering experiment

Languages Studied: Chinese, English