Automatic News Verification

EDIC candidacy exam
Exam president: Prof. Martin Jaggi
Thesis advisor: Prof. Karl Aberer
Co-examiner: Prof. Robert West

Abstract
In the new domain of data journalism, there are many approaches that address the automatic verification of news and the assessment of their credibility. In this candidacy exam we will examine three of them, namely:
    - An approach that verifies statements using an external ground-truth Knowledge Base [1]. More specifically, the authors of this paper propose a method that maps each claim (e.g., "Barack Obama is a muslim”) to a graph extracted from DBpedia. The edges of this graph have a trust value which determines the trust value of the original claim.
    - An approach that exploits information gathered manually by fact-checking websites [2]. In this paper the authors propose a platform that collects proven fake and real news by websites like PolitiFact.com and FactCheck.org and monitors how they are distributed in social networks.
    -An approach that utilizes content and social features in order to discover fake multimedia (mostly images) which are distributed in social networks [3].
At the end, we will briefly present our proposal which combines (among others) the three aforementioned approaches.

Background papers
Computational fact checking from knowledge networks, Ciampaglia, Giovanni Luca, et al. PloS one 10.6 (2015): e0128193. 
Hoaxy: A platform for tracking online misinformation, Shao, Chengcheng, et al. Proceedings of the 25th International Conference Companion on World Wide Web. International World Wide Web Conferences Steering Committee, 2016.
Challenges of computational verification in social multimedia, Boididou, Christina, et al. Proceedings of the 23rd International Conference on World Wide Web. ACM, 2014.