Bimodal fake news detection using integrated image and text analysis

Atefeh Khazaei; Gelayol Golcarenarenji; Alexander Gegov; Alexander Itchev

doi:10.1109/ICARAI70085.2026.11635523

Bimodal fake news detection using integrated image and text analysis

School of Computing, Mathematics and Physics

Research output: Chapter in Book/Report/Conference proceeding › Conference contribution

Abstract

Fake news detection is increasingly important as false information can spread quickly across the internet. This paper proposes a bimodal framework that combines text and image analysis, using XLNet for text, summarisation methods for long articles, and a lightweight ResNetRS420-based image model. Using weighted hard-voting late fusion, the approach achieved accuracies of 83.80% on GossipCop, 82.76% on PolitiFact, and 85.04% on Fakeddit, showing that combining textual and visual features improves performance while remaining computationally efficient.

Original language	English
Title of host publication	International Conference Automatics, Robotics and Artificial Intelligence, ICARAI 2026 - Proceedings
Publisher	Institute of Electrical and Electronics Engineers Inc.
Number of pages	6
ISBN (Electronic)	9798331563110
ISBN (Print)	9798331563127
DOIs	https://doi.org/10.1109/ICARAI70085.2026.11635523
Publication status	Published - 10 Aug 2026
Event	4th International Conference Automatics, Robotics and Artificial Intelligence, ICARAI 2026 - Sozopol, Bulgaria Duration: 12 Jun 2026 → 15 Jun 2026

Conference

Conference	4th International Conference Automatics, Robotics and Artificial Intelligence, ICARAI 2026
Country/Territory	Bulgaria
City	Sozopol
Period	12/06/26 → 15/06/26

Keywords

Bio modal Fake News Detection
Computer vision
Summarisation
Text analysis

Access to Document

10.1109/ICARAI70085.2026.11635523

Cite this

@inproceedings{f2af849b9b6945a3abb44b9192b18997,

title = "Bimodal fake news detection using integrated image and text analysis",

abstract = "Fake news detection is increasingly important as false information can spread quickly across the internet. This paper proposes a bimodal framework that combines text and image analysis, using XLNet for text, summarisation methods for long articles, and a lightweight ResNetRS420-based image model. Using weighted hard-voting late fusion, the approach achieved accuracies of 83.80\% on GossipCop, 82.76\% on PolitiFact, and 85.04\% on Fakeddit, showing that combining textual and visual features improves performance while remaining computationally efficient.",

keywords = "Bio modal Fake News Detection, Computer vision, Summarisation, Text analysis",

author = "Atefeh Khazaei and Gelayol Golcarenarenji and Alexander Gegov and Alexander Itchev",

note = "No ISSN; 4th International Conference Automatics, Robotics and Artificial Intelligence, ICARAI 2026 ; Conference date: 12-06-2026 Through 15-06-2026",

year = "2026",

month = aug,

day = "10",

doi = "10.1109/ICARAI70085.2026.11635523",

language = "English",

isbn = "9798331563127",

booktitle = "International Conference Automatics, Robotics and Artificial Intelligence, ICARAI 2026 - Proceedings",

publisher = "Institute of Electrical and Electronics Engineers Inc.",

address = "United States",

}

Khazaei, A , Golcarenarenji, G , Gegov, A & Itchev, A 2026, Bimodal fake news detection using integrated image and text analysis. in International Conference Automatics, Robotics and Artificial Intelligence, ICARAI 2026 - Proceedings. Institute of Electrical and Electronics Engineers Inc., 4th International Conference Automatics, Robotics and Artificial Intelligence, ICARAI 2026, Sozopol, Bulgaria, 12/06/26. https://doi.org/10.1109/ICARAI70085.2026.11635523

Bimodal fake news detection using integrated image and text analysis. / Khazaei, Atefeh ; Golcarenarenji, Gelayol ; Gegov, Alexander et al.
International Conference Automatics, Robotics and Artificial Intelligence, ICARAI 2026 - Proceedings. Institute of Electrical and Electronics Engineers Inc., 2026.

Research output: Chapter in Book/Report/Conference proceeding › Conference contribution

TY - GEN

T1 - Bimodal fake news detection using integrated image and text analysis

AU - Khazaei, Atefeh

AU - Golcarenarenji, Gelayol

AU - Gegov, Alexander

AU - Itchev, Alexander

N1 - No ISSN

PY - 2026/8/10

Y1 - 2026/8/10

N2 - Fake news detection is increasingly important as false information can spread quickly across the internet. This paper proposes a bimodal framework that combines text and image analysis, using XLNet for text, summarisation methods for long articles, and a lightweight ResNetRS420-based image model. Using weighted hard-voting late fusion, the approach achieved accuracies of 83.80% on GossipCop, 82.76% on PolitiFact, and 85.04% on Fakeddit, showing that combining textual and visual features improves performance while remaining computationally efficient.

AB - Fake news detection is increasingly important as false information can spread quickly across the internet. This paper proposes a bimodal framework that combines text and image analysis, using XLNet for text, summarisation methods for long articles, and a lightweight ResNetRS420-based image model. Using weighted hard-voting late fusion, the approach achieved accuracies of 83.80% on GossipCop, 82.76% on PolitiFact, and 85.04% on Fakeddit, showing that combining textual and visual features improves performance while remaining computationally efficient.

KW - Bio modal Fake News Detection

KW - Computer vision

KW - Summarisation

KW - Text analysis

UR - https://www.scopus.com/pages/publications/105048068218

U2 - 10.1109/ICARAI70085.2026.11635523

DO - 10.1109/ICARAI70085.2026.11635523

M3 - Conference contribution

AN - SCOPUS:105048068218

SN - 9798331563127

BT - International Conference Automatics, Robotics and Artificial Intelligence, ICARAI 2026 - Proceedings

PB - Institute of Electrical and Electronics Engineers Inc.

T2 - 4th International Conference Automatics, Robotics and Artificial Intelligence, ICARAI 2026

Y2 - 12 June 2026 through 15 June 2026

ER -

Bimodal fake news detection using integrated image and text analysis

Abstract

Conference

Keywords

Access to Document

Fingerprint

Cite this