Hack the Fake News Initiative: Machine Learning Challenge 2017

The Event

Did you know that around 20k-30k articles are published online in Bulgaria daily by news outlets, informational agencies and other informational, entertainment and niche online media?

Our initial assessment shows that among the professionally prepared news, analyses, interviews and commentaries there is a certain portion of the content that does not follow any journalism standards and worse, is a deceptive or manipulative representation of real events or an outright fabrication.

The publishers of this type of content operate multiple domains and rely on extensive popularization on social media. The average consumer hasn’t got the reflexes and the media literacy to be able to critically analyse and distinguish between proper news and the so-called “fake news” pieces.

We are turning to the power of technology to find a disruptive solution to a global trend, impacting all areas of life and public activity.

We are inviting the brightest talents in data science & programming to apply their latest knowledge in machine learning to create one or more algorithms that evaluate the possibility for a random article you encounter online to be fake.

Why You Should Participate?

• Join efforts with other experts and learn in practice more about machine learning
• Put your tech skills to a solution which impacts the whole society
• Winning teams will have the chance to make a real impact by collaboration with media, communication experts and other key players in the online environment

More About the Hackathon Challenge

Hackathon Participants will have to apply the most effective machine learning or reinforced machine learning methods to create an algorithm that looks for patterns in fake news structure, punctuation and sentence building and goes beyond occurrence of certain words or evaluation of the reliability of source or other factors that fake news publishers can easily adapt to.

With the support of media listening & analytics company Sensika we are providing participating teams with a training set of over 1.5k articles, evaluated by communication experts and a test set of more than 20 000 articles, published over a certain period of time.