Connecting the Dots Between Fact Verification and Fake News Detection

Proceedings of the 28th International Conference on Computational Linguistics, pages 1820–1825 Barcelona, Spain (Online), December 8-13, 2020 1820 Connecting the Dots Between Fact Verification and Fake News Detection Qifei Li∗ Wangchunshu Zhou∗ Beihang University {liqifei, zhouwangchunshu}@buaa.edu.cn Abstract Fact verification models have enjoyed a fast advancement in the last two years with the develop- ment of pre-trained language models like BERT and the release of large scale datasets such as FEVER. However, the challenging problem of fake news detection has not benefited from the improvement of fact verification models, which is closely related to fake news detection. In this paper, we propose a simple yet effective approach to connect the dots between fact verification and fake news detection. Our approach first employs a text summarization model pre-trained on news corpora to summarize the long news article into a short claim. Then we use a fact verification model pre-trained on the FEVER dataset to detect whether the input news article is real or fake. Our approach makes use of the recent success of fact verification models and enables zero-shot fake news detection, alleviating the need of large scale training data to train fake news detection models. Experimental results on FakenewsNet, a benchmark dataset for fake news detection, demonstrate the effectiveness of our proposed approach. 1 Introduction Recently, fake news has been appearing in large numbers, which are easily accessed and disseminated in the online world with the booming development of online social networks. Users can be affected due to the deceptive words intentionally and verifiably false, which makes fake news detection urgent and important for maintaining social order (Shu et al., 2017). Most existing methods for detecting fake news rely heavily on supervised learning on a large scale dataset with news articles labeled as fake or real by human experts. However, such a labeled dataset is difficult and time-consuming to obtain while few large scale fake news detection datasets are pub- licly available (Oshikawa et al., 2018). Meanwhile, with the fast development of pre-trained language models (Peters et al., 2018; Radford et al., 2018; Devlin et al., 2018) and the release of large scale datasets (Thorne et al., 2018), research on the fact verification task have enjoyed fast advancement (Nie et al., 2019; Hanselowski et al., 2018; Zhou et al., 2019a). However, the advancement in the field of fact verification fails to transfer to the task of fake news detection due to the different nature of input sequences (i.e., fact verification aims to check the reliability of a claim of one or a few sentences while fake news detection aims to check the trustworthy of a long article) and the lack of large scale training data to train the fact verification models for fake news detection. To address the above issues, in this paper, we propose a simple yet effective approach to connect the dots between fact verification and fake news detection. Our approach exploits off-the-shelf models in two relatively well-studied problems—text classification and fact verification—to tackle the problem of fake news …

Content truncated to 500 words