Literature Review

1. Augenstein, Isabelle, Christina Lioma, Dongsheng Wang, Lucas Chaves Lima, Casper Hansen, Christian Hansen, Jakob Grue Simonsen . MultiFC: A Real-World Multi-Domain Dataset for Evidence-Based Fact Checking of Claims. In Proceedings of the 2019 Conference on Empirical Methods in Natural Language Processing and the 9th International Joint Conference on Natural Language Processing (EMNLP-IJCNLP), pages 4677–4691, nov 2019. URL http://arxiv.org/abs/1909.03242.

   This paper presents a large real-world fact-checking dataset, which consists of 34,918 claims from 26 English fact-checking websites. Each set is accompanied by rich metadata and 10 evidence pages retrieved from Google search. Within each claim, entities such as people and places are also detected, disambiguaated, andn linked to their corresponding Wikipedia page.

2. Ding, Lixuan, Lanting Ding, and Richard O. Sinnott. “Fake News Classification of Social Media Through Sentiment Analysis.” International Conference on Big Data. Springer, Cham, 2020.

   In this paper, Sinott et al. test the hypothesis that fake news on social media typically deals with emotive topics and that a neural network equipped with sentiment classification will, therefore, perform well in fake news detection. In doing so, they train their models using Naive Bayes, Decision Tree, and bi-LSTM for sentiment classification and feature selection. News from FakeNewsNet and CredBank are employed for the algorithm’s development. Their final results demonstrate that sentiment analyses can help improve the discrimination of fake news from real news.

3. Kaliyar, Rohit Kumar, et al. “FNDNet–a deep convolutional neural network for fake news detection.” Cognitive Systems Research 61 (2020): 32-44.

   In a recent experimental project similar to our own, Kaliyar et al. test the efficacy of a deep convolutional neural network designed to automatically learn the discriminatory features for fake news classification. They benchmark their model using comprehensive datasets and consider various perfomance evaluative parameters such as Wilcoxon, false positive, and true negative to validate their results. Much to our amazement, the research group reports that their model acheived 98.36% accuracy in fake news detection.

4. Schuster, Tal, Darsh J Shah, Yun Jie Serene Yeo, Daniel Filizzola, Enrico Santus, Regina Barzilay. Towards Debiasing Fact Verification Models. In Proceedings of the 2019 Conference on Empirical Methods in Natural Language Processing and the 9th International Joint Conference on Natural Language Processing (EMNLP-IJCNLP), pages 3410–3416, 2019. ISBN 9781950737901. doi: 10.18653/v1/d19-1341. URL https://arxiv.org/pdf/1908.05267.pdf.

   In this paper, Schuster et al. addresses checking whether a claim is natural or synthetic (modified from its original version) and proposes methodology for debiasing Fact-Verification datasets. Schuster et al. proposes to synthesize each claim into a pair – one that is supported and the one that is refute dby the evidence – in order to prevent models from relying on cues from the claims. Another solution offered is an algorithm that gives more weights to claims containing give-away phrases of the opposite label to balance out their biasing effect.

5. Tan, Reuben, Kate Saenko, and Bryan A. Plummer. “Detecting Cross-Modal Inconsistency to Defend Against Neural Fake News.” arXiv preprint arXiv:2009.07698 (2020). URL: https://arxiv.org/pdf/2009.07698.pdf

   In this paper, Tan et. al. discuss the shortcomings of artificially generated news and develop a dataset with both human-written and artificially generated news articles. They then employ a visual-sementic representation of news articles based on the article text, images, and captions to train a neural network that can recognize semantic inconsistencies between recognized entities in the article images, captions and text. The improved performance of their fake news detection model based on recognizing visual-semantic inconsistencies, compared to current text-based classification models, makes them propose this as a way to combat weaknesses exploited by artificially generated news articles.

6. Le, Thai, Suhang Wang, and Dongwon Lee. “Malcom: Generating malicious comments to attack neural fake news detection models.” arXiv preprint arXiv:2009.01048 (2020). URL: https://arxiv.org/pdf/2009.01048.pdf

   In this paper, Le et. al. employ a malicious comment generation model (Malcom) as a way to exploit weaknesses in fake news detection that depends on user comments. Their model generated comments that identified different shortcomings in using classifiers that encode comments through different forms of vector representation and sequential dependency layers. At the end of their investigation, they successfully developed an adversarial model that deceives multiple state of the art fake news generation models with a 94% accuracy.

Additional Readings:

1. Hamid, Abdullah, et al. “Fake News Detection in Social Media using Graph Neural Networks and NLP Techniques: A COVID-19 Use-case.” arXiv preprint arXiv:2012.07517 (2020).

2. Oshikawa, Ray, Jing Qian, William Yang Wang. A Survey on Natural Language Processing for Fake News Detection. In Proceedings of the 12th Language Resources and Evaluation Conference, pages 6086–6093, nov 2018. URL http://arxiv.org/abs/1811.00770.

3. Thorne, James and Andreas Vlachos. Automated Fact Checking: Task Formulations, Methods and Future Directions. In Proceedings of the 27th International Conference on Computational Linguistics, pages 3346–3359, jun 2018. URL http://arxiv.org/abs/1806.07687.

4. Umer, M., Z. Imtiaz, S. Ullah, A. Mehmood, G. S. Choi and B. -W. On, “Fake News Stance Detection Using Deep Learning Architecture (CNN-LSTM),” in IEEE Access, vol. 8, pp. 156695-156706, 2020, doi: 10.1109/ACCESS.2020.3019735.