Almost three years now, Ontotext has been part of a multi-national team of scientists whose aim is to automatically detect and verify rumours (phemes) across social networks and online media. This initiative is called Pheme and is funded under the EU’s 7th FP for research, technological development and demonstration. By developing novel cross-disciplinary social semantic methods and combining them with big data analytics, Pheme will add value to many business applications such as medical information systems and digital journalism. Ontotext main contribution to this project has been to develop algorithms for rumour detection and selection of check-worthy claims in social media, thus enabling journalists to focus on tweets showing enough information to check against public sources.
How does the algorithm detect rumours?
Moreover, Ontotext’s GraphDB knowledge base has been adapted as a semantic repository with scalable lightweight reasoning, used to distinguish between rumours and facts. Datasets from Linked Open Data (LOD) cloud such as FactForge, DBpedia, OpenCyc, and Linked Life Data are the factual knowledge source for providing essential features for classifying the four kinds of rumours addressed: misinformation, disinformation, unverified information, and disputed information. On top of that, semantic annotation has been used to analyse and interlink facts, people, organisations and locations from untrusted social-media streams to the rich contextual information about them in our knowledge base. This way, we know who these people are, which organisations they are associated with, where they are based, etc.
One can freely browse and explore the Pheme database.
Thanks to our participation in this project, we are among the 40 companies selected to compete for the Innovation Radar Prize 2016. If you want to see us win, cast your vote on the website of the European Commission.