Text-image multimodal fusion model for enhanced fake news detection

Area of Science:

• Computer Science
• Artificial Intelligence
• Information Science

Background:

• The proliferation of the internet and digital technology has amplified the challenge of distinguishing authentic news from fabricated content, leading to widespread misinformation.
• Existing fake news detection methods often analyze features in isolation, neglecting the potential benefits of combining information from multiple data sources.
• Multimodal fusion offers a more robust approach by integrating diverse data types, such as text and images, for a comprehensive understanding.

Purpose of the Study:

• To develop and evaluate a novel model for fake news identification that leverages multimodal feature fusion.
• To improve the accuracy and effectiveness of fake news detection by integrating textual and visual information.
• To address the critical issue of online misinformation through an enhanced detection framework.

Main Methods:

• A framework was designed to integrate textual and visual information using early, joint, and late fusion strategies.
• Data underwent cleaning and feature extraction for both text and visual modalities before classification.
• The proposed model was trained and evaluated on the Gossipcop and Fakeddit datasets.

Main Results:

• The multimodal fusion model achieved high accuracy rates of 85% on the Gossipcop dataset and 90% on the Fakeddit dataset.
• The model demonstrated strong performance with F1-scores of 90% for Gossipcop and 88% for Fakeddit.
• Results across various training periods confirmed the efficacy of multimodal fusion in enhancing fake news detection.

Conclusions:

• Multimodal fusion, integrating text and image recognition, is a highly effective strategy for combating fake news.
• The proposed framework offers a significant advancement in misinformation detection accuracy.
• This research underscores the importance of comprehensive approaches in addressing the challenges of online fake news.