EDMO has set-up and operates a digital collaborative platform that brings together fact-checkers in a common environment, allowing them to communicate and share valuable information, and to access a set of tools, services, and resources to facilitate the detection and analysis of misinformation incidents.
The EDMO platform is based on Truly Media, a verification platform that aims to help journalists and fact-checkers to better deal with the increased volume and complexity of online disinformation. Given the fact that automated verification and fact-checking may never become possible and human judgement will always be necessary, EDMO’s platform aims to assist its users with the various fact-checking and verification tasks they perform, through the optimisation of their workflow, the facilitation of collaboration, the integration of multiple fact-checking tools, the use of analytics, and ultimately through time-saving and increased efficiency and effectiveness.
The Truly Media platform offers the following capabilities:
- A digital collaborative environment where members of the same or different organisations can connect, communicate, and work together on joint fact-checking activities/collaborative investigations;
- Tools to discover and collect content from various online sources (web, social media, online platforms) and to organise, sort and archive the collected content;
- Tools that support the analysis, verification, and fact-checking of multimedia content (text, images, videos);
- Access to data repositories (e.g. Eurostat, OECD repositories).
The platform is modular, allowing linking and API-based integration with tools/modules and services developed by third-parties, according to the technical specifications stated below.
What verification functionalities does Truly Media provide?
Advanced Search & Filtering
- In order to monitor social media content users can set up different ‘Feeds’, based on specific keywords, through a friendly and easy-to-use interface. Each feed is built based on content from a single source, mainly Facebook, Twitter, YouTube, and VK. All elements within a feed can be filtered using a variety of criteria, such as the time, source, language, and keywords, depending on the capabilities of the particular social media channel.
- Users can cross-search and explore aggregated content from different platforms. With a single search query, users can explore content from Google, Facebook, YouTube, LinkedIn, 4chan’s /pol channel, and news websites. In order to make the search easier, they are also able to use a set of advanced search filters, as well as easy-to-use Google search operators.
- Translation support (through Google Translation and CEF’s eTranslation service) is offered to make dealing with foreign language content easier.
Content Organisation, Curation & Archiving
- Truly Media supports common workspaces, called ‘Collections’, where fact checkers and collaborating journalists can define working teams to perform particular tasks on relevant content and together decide on how trustworthy the information they have found is.
- The contents of Collections are shared and visible to all team members, while changes in a Collection are visible in real-time.
- Users can add content inside a Collection from their feeds. The contents of each Collection can further be enhanced by directly importing content from Facebook, Twitter, Tweetdeck and YouTube, using a browser extension.
- Users can upload locally stored media files or paste links from social media posts or websites to add interesting content inside their Collections. All this extra content is automatically available for further investigation.
- Users can also add notes and tags to a specific item to categorize and further organise the content inside a Collection. All changes performed on Collection items are visible in real-time to all the members of the team. It is also possible to search for specific items within a Collection by applying filters.
- Since everything is organised and saved inside the platform, users can make sure that their work and findings are archived. Through a “global search” feature, users can retrieve saved content, while content inside a Collection can be exported to CSV for archiving or further analysis purposes.
- Users can work collaboratively to verify photos, videos, text, and social media accounts in real-time, by using a common workspace and a set of communication channels, like email notifications, Slack messaging, group chats, direct messages, etc.
- Collections and all individual items within a Collection are viewable by all team members. Each Collection has a dedicated chat where team members working on it can communicate. All activities that took place in a Collection are logged and can be viewed by all.
- Collaboration is a rather important part of the verification process, since users not only need to collaborate with each other but also to find and connect with other stakeholders to share their expertise or explore further collaborations. Truly Media allows users to share their Collections with other users, assign specific tasks to their team members, or find and connect with other organisations that use the platform to work together.
Digital content verification on the spot
Users can select the most important items for their story and verify each item individually using a verification checklist and integrated third-party verification tools.
With the verification checklist, users can record the results of their analysis on a specific Collection item. The checklist has been designed based on the workflow of investigative journalism. Users can also mark each item’s verification status as raw, pending, unclear, verified or fake.
Truly Media provides a variety of tools which help users determine the veracity of online content, like the following:
One of the most common issues for journalists and fact-checkers is verifying images and videos circulating on social media. With Truly Media, users can examine the authenticity of a photo or a video with:
- Reverse Image Searches: If journalists want to know an image’s origin or even find similar photos, they can search with the image URL or the file itself on the Web (in reverse image search engines like Google images or Yandex) and get information on whether the specific material has been used again, even identify near duplicate and modified images. Truly Media allows you to directly send images or video frames to Google, Yandex, and TinEye.
- Verification by location: Google Maps and Street View have been integrated to allow users to easily check locations, find the right one and save the coordinates in the item’s Checklist. You can also check weather conditions on a particular day and time in a specific location through historical weather forecasts from Wolfram Alpha.
- Image tools: Examine and analyse images through a set of functionalities like image flipping, using a magnifier, cropping or annotating parts of the image
- Video tools: Analyse videos by adding in-video annotations, extracting keyframes, sending the extracted keyframes for reverse image searching.
TruthNest is a Twitter analytics tool, which is integrated into Truly Media to assist users with verifying social media content by providing them with useful insights on an account’s online behaviour. TruthNest can prove critical at times in minimising the time from event discovery to story publication.
The tool uses several metrics in order to retrieve and provide us with a variety of analytics that can help gain additional insights about a Twitter account. More specifically, by analysing Twitter user accounts based on their activity, network and influence, TruthNest allows journalists to:
- access in-depth twitter analytics
- assess the credibility of a source
- detect trolls and bots
TruthNest’s analytics produce a set of alerts or flags which highlight suspicious behaviour and generate a bot probability score for the analysed Twitter account.
Technical Description of Truly Media
Truly Media is a web application built in Angular backed by a Java Spring boot backend. The backend is exposing a RESTful HTTP API over SSL which is consumed by the frontend on a stateless manner. The authentication is performed initially over Twitter’s OAuth2.0 mechanism. The real-time collaboration features are powered by a socket server built in socket.io/node.js. For asynchronous (heavy processing) tasks, the application follows a Service Worker pattern.
The storage layer consists of various repositories for different purposes each: A replica set NoSQL MongoDB is responsible for metadata object (JSON) storage, An Amazon S3 Bucket for binary file storage (e.g. photos/videos), Apache Solr Server for full-text indexing and retrieval and a RabbitMQ message broker for event messaging.
Most of Truly Media’s elements reside on the cloud. The back-end is running as a single process on Heroku/AWS. Several instances of this process can run on multiple Heroku Dynos exposing the same HTTP interface. The load balancing of the various instances is automatically controlled by Heroku itself, therefore the platform can offer high elasticity in terms of scalability. MongoDB is hosted on MongoLab as a replica set with automatic failover, scheduled backups and high scalability. The frontend is served by Amazon S3 and Amazon Cloudfront CDN. The Single Page Application nature of Angular lets the frontend run exclusively on the client’s browser, therefore not consuming resources of any server. The RabbitMQ message broker is provided as a service by CloudAMQP. Finally, as already mentioned, the binary/object storage is provided as a service by Amazon S3.
Technical Specifications for Integration
A prerequisite for a service in order to be integrated with Truly Media is to expose an HTTP API, more preferably following REST conventions.
Truly Media can consume the API and visualise the data provided on own User Interface. The communication with the REST interface has to be made over SSL/HTTPS and there must be an authentication mechanism (e.g. JWT).
In case a service is not providing an HTTP API, Truly Media can also request access on a dataset (e.g. Database) in order to crawl and fetch data and then feed it into the Truly Media repository. A similar approach can be followed with feeds (like RSS), which can be periodically accessed by Truly Media with scheduled tasks running offline. Finally, in case external applications or services need to integrate the Truly Media database or Truly Media’s backend services, everything is exposed as HTTP RESTful API, so it can be consumed, under the condition that the client side authenticates identity with a JWT token.