.png){kind=icon}
.png){kind=icon}
Return
A deep learning framework for detecting fake news using optimized GRU and image-text fusion
.png){kind=icon}
PRE
AI
S
M
V
B
DOI：10.1080/02533839.2026.2630738.png)
.png){kind=icon}
Original request for help.png){kind=icon}
Share.png){kind=icon}
Save
Abstract
En 中文
Fake news continues to be a growing concern across social media platforms, where users encounter a mix of text and images that can be misleading or completely false. A new deep learning framework is proposed that processes and combines both text and image information to improve fake news identification. The model uses a Gated Recurrent Unit (GRU) to analyze textual features, optimized through a metaheuristic technique called the Sparrow Search Optimizer (SSO), which fine-tunes the model's internal structure for better performance. Visual content is processed using ResNet-101, a proven convolutional neural network for extracting meaningful patterns from images. These features are then merged using Multi-Modal Bilinear Pooling (MBP), a technique that effectively combines both types of data to create a more complete representation of the news content. A softmax classifier is used at the final stage to label the content as real or fake. This hybrid model was tested using Twitter and Weibo datasets containing a wide range of real and fake news samples. The results showed a significant improvement in classification accuracy over text-only or image-only models. By integrating visual and textual elements, this system offers a more reliable solution to fake news detection in today's multimedia-driven digital landscape.
Keywords:
Fake news
text and image features
multi-modal model
optimal gated recurrent unit
multi-modal bilinear pooling
Journal
J
IF:
1.2
Papers:
122
Citations:
1.1K
Organization
No organization information available
Cited Papers