Modeling Factual Claims with Semantic Frames

Proceedings of the 12th Conference on Language Resources and Evaluation (LREC 2020), pages 2511–2520 Marseille, 11–16 May 2020 c⃝ European Language Resources Association (ELRA), licensed under CC-BY-NC 2511 Modeling Factual Claims with Semantic Frames Fatma Arslan, Josue Caraballo, Damian Jimenez, Chengkai Li Department of Computer Science and Engineering University of Texas at Arlington Arlington, TX USA {fatma.dogan, josue.caraballo, damian.jimenez}@mavs.uta.edu, cli@uta.edu Abstract In this paper, we introduce an extension of the Berkeley FrameNet for the structured and semantic modeling of factual claims. Modeling is a robust tool that can be leveraged in many different tasks such as matching claims to existing fact-checks and translating claims to structured queries. Our work introduces 11 new manually crafted frames along with 9 existing FrameNet frames, all of which have been selected with fact-checking in mind. Along with these frames, we are also providing 2, 540 fully annotated sentences, which can be used to understand how these frames are intended to work and to train machine learning models. Finally, we are also releasing our annotation tool to facilitate other researchers to make their own local extensions to FrameNet. Keywords: frame-semantics, frame-annotated corpus, factual claims, fact-checking 1. Introduction In recent years, the proliferation of misinformation has reached a staggering pace eroding people’s confidence in politics and even affected democracies (Bennett and Liv- ingston, 2018). For example, during the 2016 elections, propagated misinformation was shown to be highly in favor of one side (Allcott and Gentzkow, 2017). A recent survey1 conducted by the Pew Research Center found that 68% of respondents reported that misinformation hurt their belief in the government and 51% reported a belief that misinfor- mation can impede progress in politics. Many efforts have emerged in response to the urgent need to fight the dissem- ination of misinformation. According to a recent report2 from the Duke Reporters’ Lab, the number of active fact- checking outlets has reached 226 in 73 countries. There has also been a considerable response from the academic re- search communities within computer science, political sci- ence, and journalism. These research communities have made significant efforts in studying misinformation and in aiding fact-checking. Many such efforts led to the development of computational methods and tools in countering misinformation on various fronts, such as identifying claims worth fact-checking from a myriad of sources of digital or traditional media (Has- san et al., 2015; Patwari et al., 2017; Jaradat et al., 2018; Jimenez and Li, 2018; Hansen et al., 2019), debunking re- peated claims by matching them against a collection of al- ready checked claims3 (Hassan et al., 2017; Adair et al., 2019), and vetting claims by using supporting, refuting, and related evidence sentences from documents (Miranda et al., 2019). Several studies aimed at understanding misinfor- mation on several aspects such as its diffusion model (All- cott et al., 2019), correlations between different predictors and an individual’s tendency to reject or accept a factual claim (Pennycook and Rand, 2019), the effects of different 1 https://pewrsr.ch/37ykPcs 2 https://reporterslab.org/tag/fact-checking-database/ 3 Duke Reporters’ Lab in-house fact-checking app …

Content truncated to 500 words