Words are the Window to the Soul: Language-based User Representations for Fake News Detection

Proceedings of the 28th International Conference on Computational Linguistics, pages 5467–5479 Barcelona, Spain (Online), December 8-13, 2020 5467 Words are the Window to the Soul: Language-based User Representations for Fake News Detection Marco Del Tredici Amazon mttredic@amazon.com Raquel Fern´andez University of Amsterdam raquel.fernandez@uva.nl Abstract Cognitive and social traits of individuals are reflected in language use. Moreover, individuals who are prone to spread fake news online often share common traits. Building on these ideas, we introduce a model that creates representations of individuals on social media based only on the language they produce, and use them to detect fake news. We show that language-based user representations are beneficial for this task. We also present an extended analysis of the language of fake news spreaders, showing that its main features are mostly domain independent and consistent across two English datasets. Finally, we exploit the relation between language use and connections in the social graph to assess the presence of the Echo Chamber effect in our data. 1 Introduction Fake news have become a problem of paramount relevance in our society, due to their large diffusion in public discourse, especially on social media, and their alarming effects on our lives (Lazer et al., 2018). Several works show that fake news played a role in major events such as the US Presidential Elections (Allcott and Gentzkow, 2017), stock market trends (Rapoza, 2017), and the Coronavirus disease outbreak (Shimizu, 2020). In NLP a considerable amount of work has been dedicated to fake news detection, i.e., the task of classifying a news as either real or fake – see Zhou and Zafarani (2020), Kumar and Shah (2018) and Oshikawa et al. (2020) for overviews. While initial work focused uniquely on the textual content of the news (Mihalcea and Strapparava, 2009), subsequent research has considered also the social context in which news are consumed, characterizing, in particular, the users who spread news in social media. In line with the results reported in other classification tasks of user-generated texts (Del Tredici et al., 2019; Pan and Ding, 2019), several studies show that leveraging user representations, together with news’ ones, leads to improvements in fake news detection. In these studies, user representations are usually computed using informative but costly features, such as manually assigned credibility scores (Kirilin and Strube, 2018). Other studies, which leverage largely available but scarcely informative features (e.g., connections on social media), report less encouraging results (Zubiaga et al., 2016). Our work also focuses on users. We build on psychological studies that show that some people are more prone than others to spread fake news, and that these people usually share a set of cognitive and social factors, such as personality traits, beliefs and ideology (Pennycook et al., 2015; Pennycook and Rand, 2017). Also, we rely on studies showing a relation between these factors and language use, both in Psychology and Linguistics (Pennebaker et al., 2003; De Fina, 2012) and in NLP (Plank and Hovy, 2015; Preot¸iuc-Pietro et al., 2017). We therefore propose to leverage user-generated …

Content truncated to 500 words