RCED: Relational Graph Convolutional Network with Commonsense Knowledge for Emotion Recognition and Depression Early Warning in Dialogues
Summary
No summary yet. VIRAT can read this paper and write a plain-language summary of its aim, methods, key findings and limitations.
Summarise with VIRAT Free to try with your Science Catalogs account.Abstract
In this paper, we propose a detailed graph neural network approach to the task of ERC, RCED, Relational graph convolutional network with Commonsense knowledge for Emotion recognition and Depression early warning in dialogues. We infer specific emotional labels for each dialogue in a discussion by evaluating the overall conversational context combined with interspeaker dynamics. The model incorporates a robust DeBERTa feature extractor, utilizes a relational graph structure, and employs large language models for advanced commonsense reasoning, addressing key limitations of previous ERC methods. Experiments conducted on the IEMOCAP dataset demonstrate the overall effectiveness of our approach. By integrating psychological principles into our model’s applications, we aim to extend its contributions to a societal context, particularly in online conversational settings. Leveraging predicted emotional labels in ERC, we infer psychological factors such as pleasure, arousal, and emotional consistency to enable the early detection and prevention of depression.
Fields of Study & Subjects
Emotion and Mood Recognition
Sentiment Analysis and Opinion Mining
Mental Health via Writing
Publication details
- Journal / source
- IEEE International Conference on Power, Electronics and Green Energy (ICPEGE)
- Pages
- 265–270
- Publication date
- 28 January 2026
- Language
- EN