Rhetorical Structure Approach for Online Deception Detection: A Survey

Proceedings of the 13th Conference on Language Resources and Evaluation (LREC 2022), pages 5906–5915 Marseille, 20-25 June 2022 © European Language Resources Association (ELRA), licensed under CC-BY-NC-4.0 5906 Rhetorical Structure Approach for Online Deception Detection: A Survey Francielle Vargas*†, Jonas D’Alessandro∝, Zohar Rabinovich‡ Fabr´ıcio Benevenuto†, Thiago A. S. Pardo* *Institute of Mathematical and Computer Sciences, University of S˜ao Paulo, Brazil ‡Viterbi School of Engineering, University of Southern California, United States †Computer Science Department, Federal University of Minas Gerais, Brazil ∝Linguistics Department, Federal University of Minas Gerais, Brazil francielleavargas@usp.br, jonasd@let.grad.ufmg.br, zoharrab@usc.edu fabricio@dcc.ufmg.br, taspardo@icmc.usp.br Abstract Most information is passed on in the form of language. Therefore, research on how people use language to inform and misinform, and how this knowledge may be automatically extracted from large amounts of text is surely relevant. This survey provides first-hand experiences and a comprehensive review of rhetorical-level structure analysis for online deception detection. We systematically analyze how discourse structure, aligned or not with other approaches, is applied to automatic fake news and fake reviews detection on the web and social media. Moreover, we categorize discourse-tagged corpora along with results, hence offering a summary and accessible introductions to new researchers. Keywords: rhetorical structure theory, discourse-level structure analysis, online deception detection, natural language processing. 1. Introduction Assessing whether reported statements are intention- ally misstated (or manipulated) is of considerable in- terest to researchers, financial companies, security, and governmental regulators (Larcker and Zakolyuk- ina, 2012; Larcker and Zakolyukina, 2012). Accord- ing to Hancock and Guillory (2015), there are reliable cues for deception detection, and the belief that liars give cues that may indicate their deception is nearly universal. Moreover, the most relevant literature on deception suggest that liars may be identified by their words (Newman et al., 2003; DePaulo et al., 2003), and a fairly straightforward element for mitigating risks of deceptive activities is to identify deceptive intentions (Ho and Hancock, 2019). For the last few years, there has been a growth in the number of web and social media users; consequently, the potential for deceptive activities has also increased, such as fake news, fake reviews (also known as opinion spam), deceptive discussion, and simple lies. Particu- larly, fake news detection is defined as the prediction of the chances of a news article being intentionally decep- tive (Rubin et al., 2015). Fake reviews - also known as opinion spam - are inappropriate or fraudulent reviews (Li et al., 2014; Ott et al., 2011). Deceptive discussions consist of intentionally misstated (or manipulated) nar- ratives or statements (Larcker and Zakolyukina, 2012). Notwithstanding the lack of discourse processing for deceptive detection, more recently, discourse frame- works such as Rhetorical Structure Theory (RST) (Mann and Thompson, 1987; Thompson and Mann, 1988) have been adopted to automatically detect decep- tive stories. Figure 1 shows an example of a deceptive story (fake news) annotated using the RST framework. Figure 1: Discourse-tagged fake news using RST framework. This fake news was extracted from Poli- tifact stated on April 24, 2022 in a Instagram post. As shown …

Content truncated to 500 words