A Deep Learning Approach for Automatic Detection of Fake News

Research Paper Published at ICON 2019, indexed in ACL Anthology by Tanik Saikh, Arkadipta De, Asif Ekbal, Pushpak Bhattacharyya

Proceedings

16th International Conference on Natural Language Processing 2021 - https://ltrc.iiit.ac.in/icon2019/icon2019proceedings.pdf

Introduction

Fake news detection is a very prominent and essential task in the field of journalism. This challenging problem is seen so far in the field of politics, but it could be even more challenging when it is to be determined in the multi-domain platform. In this paper, we propose two effective models based on deep learning for solving fake news detection problem in online news contents of multiple domains. We evaluate our techniques on the two recently released datasets, namely FakeNews AMT and Celebrity for fake news detection. The proposed systems yield encouraging performance, outperforming the current handcrafted feature engineering based state-of-theart system with a significant margin of 3.08% and 9.3% by the two models, respectively. In order to exploit the datasets, available for the related tasks, we perform cross-domain analysis (i.e. model trained on FakeNews AMT and tested on Celebrity and vice versa) to explore the applicability of our systems across the domains.

Result

Dataset	System Model	Test Accuracy(%)
FakeNews AMT	Proposed Model1	77.08
FakeNews AMT	Proposed Model2	83.33
FakeNews AMT	(Perez-Rosas et al. 2018) Linear SVM	74
Celebrity	Proposed Model1	76.53
Celebrity	Proposed Model2	79
Celebrity	(Perez-Rosas et al. 2018) Linear SVM	76

Reads

Read the paper at : https://arxiv.org/abs/2005.04938

Citation

For Research Puropose cite the following:

@article{DBLP:journals/corr/abs-2005-04938,
  author    = {Tanik Saikh and
               Arkadipta De and
               Asif Ekbal and
               Pushpak Bhattacharyya},
  title     = {A Deep Learning Approach for Automatic Detection of Fake News},
  journal   = {CoRR},
  volume    = {abs/2005.04938},
  year      = {2020},
  url       = {https://arxiv.org/abs/2005.04938},
  archivePrefix = {arXiv},
  eprint    = {2005.04938},
  timestamp = {Thu, 14 May 2020 16:56:02 +0200},
  biburl    = {https://dblp.org/rec/journals/corr/abs-2005-04938.bib},
  bibsource = {dblp computer science bibliography, https://dblp.org}
}