Interpretable Propaganda Detection in News Articles

Proceedings of Recent Advances in Natural Language Processing, pages 1597–1605 Sep 1–3, 2021. https://doi.org/10.26615/978-954-452-072-4_179 1597 Interpretable Propaganda Detection in News Articles Seunghak Yu1∗ Giovanni Da San Martino2 Mitra Mohtarami3 James Glass3 Preslav Nakov4 1 Amazon Alexa AI, Seattle, WA, USA 2 Department of Mathematics, University of Padova, Italy 3 MIT Computer Science and Artificial Intelligence Laboratory, Cambridge, MA, USA 4 Qatar Computing Research Institute, HBKU, Qatar yuseungh@amazon.com, dasan@math.unipd.it {mitra, glass}@csail.mit.edu, pnakov@hbku.edu.qa Abstract Online users today are exposed to mislead- ing and propagandistic news articles and me- dia posts on a daily basis. To counter thus, a number of approaches have been designed aiming to achieve a healthier and safer online news and media consumption. Automatic sys- tems are able to support humans in detecting such content; yet, a major impediment to their broad adoption is that besides being accurate, the decisions of such systems need also to be interpretable in order to be trusted and widely adopted by users. Since misleading and pro- pagandistic content influences readers through the use of a number of deception techniques, we propose to detect and to show the use of such techniques as a way to offer interpretabil- ity. In particular, we define qualitatively de- scriptive features and we analyze their suitabil- ity for detecting deception techniques. We fur- ther show that our interpretable features can be easily combined with pre-trained language models, yielding state-of-the-art results. 1 Introduction With the rise of the Internet and social media, there was also a rise of fake (Nguyen et al., 2020), biased (Baly et al., 2020a,b), hyperpartisan (Potthast et al., 2018), and propagandistic content (Da San Martino et al., 2019b). In 2016, news got weaponized, aim- ing to influence the US Presidential election and the Brexit referendum, making the general public concerned about the dangers of the proliferation of fake news (Howard and Kollanyi, 2016; Faris et al., 2017; Lazer et al., 2018; Vosoughi et al., 2018; Bovet and Makse, 2019). There ware two reasons for this. First, disinfor- mation disguised as news created the illusion that the information is reliable, and thus people tended to lower their barrier of doubt compared to when information came from other types of sources. ∗ Work conducted while the author was at MIT CSAIL. Figure 1: Comparison of propaganda prediction in- terpretability using existing methods. Our proposed method helps users to interpret propaganda predictions across various dimensions, e.g., is there a lot of pos- itive/negative sentiment (can signal the use of loaded language, which appeals to emotions), are the target sentence and the document body related to the title, does the sentence agree/disagree with the title, etc. Each symbol in the top bar chart represents an infor- mation source for propaganda detection. Second, the rise of citizen journalism led to the proliferation of various online media, and the ve- racity of information became an issue. In practice, the effort required to fact-check the news, and its bias and propaganda remained the same or even got more complex, compared to traditional …

Content truncated to 500 words