DFKI-DKT at SemEval-2017 Task 8: Rumour Detection and Classification using Cascading Heuristics

Abstract

We describe our submissions for SemEval-2017 Task 8, Determining Rumour Veracity and Support for Rumours. The Digital Curation Technologies (DKT) Sasaki, 2016, 2015) team at the German Research Center for Artificial Intelligence (DFKI) participated in two subtasks: Subtask A (determining the stance of a message) and Subtask B (determining veracity of a message, closed variant). In both cases, our implementation consisted of a Multivariate Logistic Regression (Maximum Entropy) classifier coupled with hand-written patterns and rules (heuristics) applied in a post-process cascading fashion. We provide a detailed analysis of the system performance and report on variants of our systems that were not part of the official submission.

Simple Open Stance Classification for Rumour Analysis

“…Half of the systems employed ensemble classifiers, where classification was obtained through majority voting (Wang et al, 2017;García Lozano et al, 2017;Bahuleyan and Vechtomova, 2017;Srivastava et al, 2017). In some cases the ensembles were hybrid, consisting both of machine learning classifiers and manually created rules, with differential weighting of classifiers for different class labels (Wang et al, 2017;García Lozano et al, 2017;Srivastava et al, 2017).…”

An Infrastructure for Empowering Internet Users to Handle Fake News and Other Online Media Phenomena

“…Humanity is transitioning into becoming a digital society, or at least a "digital first" society, i.e., news, media, facts, rumours (Zubiaga et al 2016;Derczynski et al 2017;Srivastava et al 2017), information are created, circulated and consumed online. Already now the right social media strategy can make or break an election or influence if a smaller or larger societal or demographic group (city, region, country, continent) is in favour or against constructively solving a certain societal challenge.…”

From Clickbait to Fake News Detection: An Approach based on Detecting the Stance of Headlines to Articles

“…The first, binary classification of "related" vs. "unrelated" can be exploited for clickbait detection. The more fine-grained classification of related headlines can specifically support in the detection of political bias and rumour veracity (Srivastava et al, 2017).…”