Analyzing and verifying digital content with the help of digital technology has kept us busy for quite some time. Our first (EU-funded) project that dealt with the matter was REVEAL. The project focused on text and image analysis and ran from late 2013 until the end of 2016. It laid the groundwork for other efforts in the verification domain – like Truly Media, our platform for collaborative verification. Truly was developed in a public-private partnership with the Athens Technology Center (ATC). While a large chunk of development costs were covered by the Google DNI Fund, much of the concept behind Truly Media is based on learnings and results coming out of REVEAL. More in-depth discussions and frequent tests with journalist from both inside and outside DW also played a big part in the platform development process. Eventually, all the know-how found its way into the final product that is now available as a commercial service.
The next (EU-funded) project that dealt with algorithm-supported verification was InVID. InVID ran from January 2016 until December 2018. It focused on the highly challenging task of partly automated video analysis/verification. Some of the results made it to prototype status only, while other outcomes have been available to (and used by) the public since mid-2017. A big success is the InVID Verification plug-in, which was specifically designed to support journalists, human rights workers, and activists. As of January 2019, the free browser extension has been downloaded by more than 8.000 individuals and received a lot of positive feedback.
Just in time, and once again working with funds from the EU’s H2020 program, we have now started another project that deals with digital content analysis, verification, and the debunking of misleading and manipulated material: WeVerify. This one will run until late 2021 and aims to address content verification challenges through a participatory verification approach, using open source algorithms, low-overhead human-in-the-loop machine learning and intuitive visualizations. Social media and web content will be analysed and contextualised within the broader online ecosystem. We will try and expose fabricated content via cross-modal content verification, social network analysis, micro-targeted debunking, and a blockchain-based public database of known fakes. The WeVerify platform for collaborative, decentralised content verification, tracking, and debunking is an excepted key outcome.
DW’s partners in the WeVerify effort are ontotext (= the overall project coordinator), the University of Sheffield (= the technical coordinator); Agence France Press (AFP), ATC, the Centre for Research & Technology Hellas (CERTH), and the EU DisinfoLab.