While working on the WeVerify H2020 funded project, we create tools helping journalists and fact-checkers fight disinformation. One aspect of our work is creating a database of known fakes which can help minimize the time to identify disinformation which has already been debunked in other context or location. Another useful and innovative tool is a verification companion that will help both professionals and citizens find the best tool or information to help them decide whether information they come across in social media is true or a fake. We plan to combine both these aspects and see how their synergies can help navigating the COVID-19 ocean of misinformation.
The problem we tackle
In addition to many other challenges, the COVID-19 crisis has given rise to so much disinformation published and shared online, that the society is facing a “disinfodemic”. Daily, tens of millions of posts about COVID-19 are shared through Facebook, Twitter, Instagram, and other platforms and journalists, fact-checkers and health authorities are overwhelmed by the effort to monitor and debunk them effectively. The International Fact-Check Network has over 100 fact-checkers working daily in over 70 countries, who have so far debunked over 3,500 false stories, however this is just a drop in the ocean.
Our solution is based on being able to automatically process a claim which appears in a tweet and allow the user to quickly determine whether it contains disinformation. While the final judgement is made by the user, we will recommend appropriate forensics tools to be applied to the associated image or video, as well as enable searching online or within a database of IFCN debunks for similar content. This will significantly speed up the process of debunking and spare valuable time in debunking the same or similar claims over and over again.
Impact on the COVID-19 crisis
Our solution helps both professionals (journalists and fact checkers) and regular citizens to quickly check a tweet about COVID-19 and see whether similar disinformation has already been debunked. If not, the verification assistant will recommend appropriate tools available as part of the Verification plug-in link to help analyse whether an image or video has been tampered with.
The value of our solution after the crisis
In recent years, even before the COVID-19 crisis, disinformation claims have been constantly flooding social media. Both professional fact checkers and regular citizens will benefit from having a quick way to consult whether particular claim has already been debunked or be advised on the appropriate tools to analyse shared multi-media and see if it has been tampered with.
What we achieved during the hackathon
During the EUvsVirus hackathon we successfully implemented functionalities that:
- Extract the text of a tweet, as well as image or video links
- Recommend to the user the most appropriate tool from the Verification plug-in toolset
- Post queries to the database of already debunked fakes to retrieve similar debunks and present them to the user
- Parse and load into the database of known fakes hundreds of debunks related to COVID-19 from one of the IFCN sources
Challenges we ran into
Our solution required addressing the challenge of parsing debunking posts in order to be able to extract from them the actual claim, where it appears (social media or elsewhere) and different evidences. Once we identify all these parts correctly, we make them searchable and accessible in the database of known fakes. Meanwhile, we also had to parse the tweets to be able to correctly separate the text from media and offer respective tools to the user of the verification assistant.
The necessities in order to continue the project
Our work will continue as part of the WeVerify project, funded by the H2020 programme of the EC. However, continuously adding new sources of debunks to the database requires more resources than we have secured in WeVerify. Finding ways to further support our efforts will ensure we can cover more IFCN sources.
What's next for Database of known fakes and verification assistant
We plan to add more functionalities to the database and to the verification assistant as well as to continue work on their integration to allow users use the full set of search queries. In order to provide diverse searches, we will also continue our work on parsing texts of the debunks in order to extract more metadata, e.g. author and dates. The verification assistant will be further developed to be able to parse not only tweets but other types of social media posts as well.