Detecting fake news with help of Event Registry

Event Registry’s news feed can be tapped for many uses. One is to help detect fake news, as recently demonstrated by a group of students from the US.

Rise of fake news

Fake news has become a central part of what is known as the post-truth political era. It is defined by political discourse appealing to people’s emotions, even when that means blatant disregard for the facts and the truth.

A screenshot taken of a fake news article published on a site that ceased to exist soon after the 2016 US election. Source

The term ‘fake news’ gained worldwide prominence during the 2016 US presidential campaign, when dozens of stories by unknown media outlets carrying misleading and false claims went viral. It has since been used in various contexts to refer to anything ranging from propaganda to outright lies. Additionally, it is frequently tossed around by politicians to attack critical media reporting.

Fighting the lies

The post-truth political culture and the associated epidemic of false stories related to the US election has prompted push back by traditional media outlets and others who fear its effect on democratic discourse. Fact checking and other efforts to identify false stories and dodgy sites have quickly become a priority.

As part of the response, various tools have emerged attempting to identify fake news. Many are designed to fact-check stories or allow users to report questionable assertions. Others provide data on the reputation of the website carrying the report.

A screenshot of the Decodex application built by French paper Le Monde to check the reputation of news websites.

Meanwhile, a group of students from Swarthmore College, in the US, recently built a fake news detection application which seeks to use stance detection technology. It enables assertions to be compared to how reputable media sources are reporting on the issue. To achieve this, they tapped Event Registry’s database.

How can Event Registry help

The Fake Bananas fake news detector allows users to enter a claim which is then checked against articles in Event Registry’s newsfeed, explains Henry Han, one of the four Swarthmore sophomores working on the project.

A screenshot of the Fake Bananas app.

The articles are then run through a stance detection machine learning model developed by the students to determine the stance of reputable sources on the inputted claim. The machine lets the user know if the claim is supported or not by the sources.

The Fake Bananas project finished top 10 at HackMIT held at the Massachusetts Institute of Technology in September, winning awards for ‘best use of data’ and ‘best use of machine learning for the common good’.

Plugging into the ER newsfeed

Fake news detection is just one of many potential applications of Event Registry’s news feed. With over 30,000 sources from around the world tracked in real-time, it provides an extremely rich set of news data which can be tapped via API or ER’s native web application.

API access to the database helps power applications in areas ranging from risk management, stock trading, market and sentiment analysis, news bias detection and many others.

Gorazd Jukovic


Leave A Comment