|
Title:
|
FAKE NEWS DETECTION: GAN-BASED DATA AUGMENTATION APPROACH |
|
Author(s):
|
Eya Maazoun, Amal Rekik and Salma Jamoussi |
|
ISBN:
|
978-989-8704-71-9 |
|
Editors:
|
Paula Miranda and Pedro Isaías |
|
Year:
|
2025 |
|
Edition:
|
Single |
|
Keywords:
|
Fake News Detection, Embeddings, BERT, GAN, Data Augmentation, Text Classification |
|
Type:
|
Full Paper |
|
First Page:
|
67 |
|
Last Page:
|
74 |
|
Language:
|
English |
|
Cover:
|
|
|
Full Contents:
|
if you are a member please login 
|
|
Paper Abstract:
|
Social media has made it easier for fake news to spread quickly, making it hard to tell which information is true or false.
While machine learning techniques have shown promise in automating fake news detection, their effectiveness is often
limited by small or imbalanced datasets. In this paper we propose a novel approach that leverages Generative Adversarial
Networks (GANs) for data augmentation to enhance fake news classification. For this purpose, we use BERT to generate
contextual embeddings at both word and phrase levels, and train GANs to produce realistic synthetic embeddings. These
synthetic samples are then combined with the original dataset to train a binary classifier. Experimental results on the
LIAR dataset demonstrate that augmenting the training data with GAN-generated embeddings significantly improves
classification performance compared to models trained without augmentation. |
|
|
|
|
|
|
