Leveraging Knowledge Graphs for CheapFakes Detection: Beyond Dataset Evaluation

Dao, Minh-Son; Zettsu, Koji

doi:10.1109/icmew59549.2023.00024

Cited by 3 publications

(3 citation statements)

References 21 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…• Knowledge graph [16], an approach that utilizes knowledge graph to learn more information from external knowledge. We use four common metrics: F1-score, precision, recall, and accuracy.…”

Section: Resultsmentioning

confidence: 99%

“…Spotfake [50] 0.5350 0.5252 0.5306 0.5279 Self-Query [14] 0.6106 0.5907 0.6320 0.5970 EANN [51] 0.6073 0.6025 0.6122 0.6300 MUICG [52] 0.7940 0.7885 0.7995 0.7930 COSMOS [1] 0.8100 0.8520 0.7720 0.8190 Knowledge graph [16] 0.8405 0.8407 0.8405 0.8405 CSGM (ours) 0.8569 0.8460 0.8680 0.8550 NLI models have little impact on the prediction outcomes.…”

Section: Methods F1-score Precision Recall Accuracymentioning

confidence: 99%

“…Detecting out-of-context misinformation has been extensively researched [1,[6][7][8], and sustained efforts have been made to develop solutions for it [9][10][11][12][13][14][15][16][17]. Many datasets have been developed for this problem, including the MAIM [18], Twitter-COMMs [19], NewsCLIPpings [20] and the COSMOS dataset.…”

Section: Cheapfakes Detectionmentioning

confidence: 99%

See 2 more Smart Citations

Multimodal scene-graph matching for cheapfakes detection

Nguyen,

Dao

et al. 2024

Preprint

View full text Add to dashboard Cite

The development of technology and social media platforms has led to the proliferation of fake news, including the cheapfakes problem. Cheapfakes can be produced easily and spread quickly; a common type is out-of-context misinformation. It is worth investigating a proper algorithm that can efficiently detect such types of misinformation. In this work, we study a new approach to the detection problem of cheapfakes by using graphical neural networks to detect out-of-context samples using the dataset from the ICME 23 Grand Challenge on Detecting Cheapfakes. Specifically, our model utilizes scene graph matching and a language model pre-trained for the natural language inference task to solve task 1 of the challenge. We also propose effective methods to generate labeled data from an unlabeled training set of the challenge. Our proposed method achieved an F1-score of 85.69% and an accuracy score of 85.50% on the public testing set, surpassing the baseline by 4.69% and 3.60%, respectively.

show abstract

Section: Resultsmentioning

confidence: 99%