EUVIMEDEUVIMED entdecken
Open Access · CC BY 4.0

Emotion-RGC net: A novel approach for emotion recognition in social media using RoBERTa and Graph Neural Networks

Jiangting Yan, Pengju Pu iD, Liheng Jiang iD

PLOS ONE · 2025

DOI: 10.1371/journal.pone.0318524 ↗

Kurz gefasst

Fachartikel in der Zeitschrift PLOS ONE, 2025. Verfasst von Jiangting Yan und 2 weiteren Autor:innen. Beteiligte Einrichtung: School of Art & Design, Shaanxi Fashion Engineering University, Xi'an, China. Hauptthemen laut MEDLINE: Social Media, Emotions, Neural Networks, Computer. Die Arbeit wurde bisher 23-mal zitiert (Crossref) und nennt 43 Literaturangaben. Der Volltext ist frei zugänglich (Open Access, CC BY 4.0). Gefördert von: 2022 Project of Shaanxi Province’s 14th Five Year Plan for Education and Science: Research on the Training.

AbstractSchlagwörterBibliografische AngabenZitierenVerwandte Studien

Abstract (Originalsprache)

Emotion recognition in social media is a challenging task due to the complex and unstructured nature of user-generated content. In this paper, we propose Emotion-RGC Net, a novel deep learning model that integrates RoBERTa, Graph Neural Networks (GNN), and Conditional Random Fields (CRF) to enhance the accuracy and robustness of emotion classification. RoBERTa is employed for effective feature extraction from unstructured text, GNN captures the propagation and influence of emotions through user interactions, and CRF ensures global consistency in emotion label prediction. We evaluate the proposed model on two widely-used datasets, Sentiment140 and Emotion, demonstrating significant improvements over traditional machine learning models and other deep learning baselines in terms of accuracy, recall, F1-score, and AUC. Emotion-RGC Net achieves an accuracy of 89.70% on Sentiment140 and 88.50% on Emotion, highlighting its effectiveness in handling both coarse- and fine-grained emotion classification tasks. Despite its strong performance, we identify areas for future research, including reducing the model’s reliance on large labeled datasets, improving computational efficiency, and incorporating temporal dynamics to capture emotion evolution in social networks. Our results indicate that Emotion-RGC Net provides a robust solution for emotion recognition in diverse social media contexts.

Schlagwörter

MeSH-Schlagwörter (MEDLINE)

Social MediaEmotionsNeural Networks, ComputerHumansDeep LearningMachine LearningGraph Neural Networks

Autor:innen und Institutionen

  • School of Art & Design, Shaanxi Fashion Engineering University, Xi'an, China.
  • CVS Health Corporation, Woonsocket, Rhode Island, USA.

ORCID-Profile: Pengju Pu, Liheng Jiang

Förderung

  • 2022 Project of Shaanxi Province’s 14th Five Year Plan for Education and Science: Research on the Training

Zitieren

APA

Yan, J., Pu, P., & Jiang, L. (2025). Emotion-RGC net: A novel approach for emotion recognition in social media using RoBERTa and Graph Neural Networks. PLOS ONE. https://doi.org/10.1371/journal.pone.0318524

Vancouver

Yan J, Pu P, Jiang L. Emotion-RGC net: A novel approach for emotion recognition in social media using RoBERTa and Graph Neural Networks. PLOS ONE. 2025. doi:10.1371/journal.pone.0318524

BibTeX
@article{Yan2025,
  author = {Yan, Jiangting and Pu, Pengju and Jiang, Liheng},
  title = {Emotion-RGC net: A novel approach for emotion recognition in social media using RoBERTa and Graph Neural Networks},
  journal = {PLOS ONE},
  year = {2025},
  doi = {10.1371/journal.pone.0318524}
}
Kostenlos starten →