Leveraging GANs and Large Language Models for Early Detection of Rumors in Persian Tweets

Document Type : Research Article

Authors

1 Department of Software Engineering University of Isfahan Isfahan, Iran

2 Department of Software Engineering, University of Isfahan, Isfahan, Iran

10.22108/jcs.2026.148800.1202

Abstract

Social networks play a crucial role in communication between individuals, information exchange, and message dissemination, making them one of the primary sources of news consumption worldwide. In recent years, social networks have been increasingly used for spreading rumors. Rumors cover various domains, including politics, society, culture, and tourism, and can inflict irreversible harm on both individuals and society. Therefore, it is essential to respond to rumors as quickly as possible and take measures to prevent their spread. If a rumor is identified within the first few hours, it is more likely that its negative and harmful effects can be prevented. Most studies on rumor detection have primarily focused on achieving high accuracy and often rely on large volumes of data related to content, network structure, and dissemination patterns. This paper proposes an early rumor detection model through data augmentation and generation of texts similar to original samples using the Generative Adversarial Network (GAN) approach and Large Language Model. The main novelty of this work is in its potential for early rumor detection through rapidly building the required detection model for each specific family of rumors within a short time frame after sensing the start of rumor spreading process. Experimental evaluation results on a Dataset of Persian Tweets indicate that employing the proposed model not only can discover the rumors with 83.88% accuracy and 87.9% F1-score.This study is a step towards controlling rumor propagation, and can greatly help improving the information quality and sensitivity of public awareness.

Keywords

Main Subjects



Articles in Press, Accepted Manuscript
Available Online from 18 November 2026
  • Receive Date: 04 May 2026
  • Revise Date: 30 October 2026
  • Accept Date: 18 November 2026