DeClarE: Debunking Fake News and False Claims using Evidence-Aware Deep Learning

Abstract

Misinformation such as fake news is one of the big challenges of our society. Research on automated fact-checking has proposed methods based on supervised learning, but these approaches do not consider external evidence apart from labeled training instances. Recent approaches counter this deficit by considering external sources related to a claim. However, these methods require substantial feature modeling and rich lexicons. This paper overcomes these limitations of prior work with an end-toend model for evidence-aware credibility assessment of arbitrary textual claims, without any human intervention. It presents a neural network model that judiciously aggregates signals from external evidence articles, the language of these articles and the trustworthiness of their sources. It also derives informative features for generating user-comprehensible explanations that makes the neural network predictions transparent to the end-user. Experiments with four datasets and ablation studies show the strength of our method. 1 As fully objective and unarguable truth is often elusive or ill-defined, we use the term credibility rather than "truth".

False News On Social Media

“…A [47], who build a framework to classify true and false claims, and also provide self-evidence for the credibility assessment. They evaluate their model against some state-of-the-art techniques on different collections of news articles (cf.…”

“…They also apply the Elaboration Likelihood Model[41] to news categories, and suggest that consuming false news requires little energy and cognition, making them more appealing to the readers. A neural network model is also presented byPopat et al (2018)…”

Explainable Fact Checking with Probabilistic Answer Set Programming

“…Given a KG K and a claim f , several approaches have been developed to estimate if f is a valid claim in K. In some of these methods, facts in the KG are leveraged to create features, such as paths [20,4] or embeddings [2,22], which are then used by classifiers to label as true or false a given test claim. Other methods rely on searching for occurrences of the given claim on Web pages [5,18]. However, such models are based on Machine Learning (ML) classifiers that in the best case can report the source of evidence for a decision but lack the ability to provide comprehensible descriptions of how a decision has been taken for a given claim.…”

“…For these reasons, in many cases, rules cannot be triggered. We identify these cases and resort to mining Web pages to get evidence for missing facts that are crucial to reach a decision for a claim [18].…”

Learning from Fact-checkers

“…To fight against fake news, many fact-checking systems ranging from human-based systems (e.g. Snopes.com), classical machine learning frameworks [20,34,38] to deep learning models [29,39,56,57] were developed to determine credibility of online news and information. However, falsified news is still disseminated like wild fire [31,59] despite dramatic rise of fact-checking sites worldwide [21].…”