On Creation of Synthetic Samples from GANs for Fake News Identification Algorithms

• Type: Article in International Conference Proceedings Book
• Date: 2022

Authors

Vaz, B (Author) Other; Bernardes, V (Author) Other; Figueira, A (Author) FCUP

Conference proceedings International

Title: Lecture Notes in Networks and Systems

Pages: 316-326

10th World Conference on Information Systems and Technologies, WorldCIST 2022

12 April 2022 through 14 April 2022

Indexing

Scopus - 0 Citations

Other information

Authenticus ID: P-00W-GTM

DOI: 10.1007/978-3-031-04829-6_28

Abstract (EN): The use of Generative Adversarial Networks is almost traditional in creating synthetic images for medical purposes. They are probably the best use of GANs until now, as their results can easily be checked by the eye of specialists. In fake news detection models, we have seen lately that neural models (and deep learning) can provide a considerable improvement from standard classifiers. Yet, the most problematic problem still is the lack of data, mostly fake news data to feed these models. In this paper, we address that by proposing the use of a GAN. Results show a better capacity to generalize when used for training an extended dataset based on synthetic samples created by this GAN. © 2022, The Author(s), under exclusive license to Springer Nature Switzerland AG.

Language: English

Type (Professor's evaluation): Scientific