Fakebox analyzes news articles to assess whether they are likely to be real news or not. By looking at a range of available aspects of an article (title, content and URL) using built-in machine learning models and a manually curated database, Fakebox can successfully identify fake news.

• BLOG: I trained fake news detection AI with >95% accuracy, and almost went crazy

Uses for Fakebox

This capability has a variety of utilities:

• Warn users before they share questionable content on your platform
• Ensure your article content is impartial and your titles are not clickbait
• Analyze ad placement to determine how often ads are placed on questionable sites
• Give your manual content moderation a head start or helping hand

What Fakebox checks

Fakebox checks the following aspects of an article. Users must provide at least one property to be checked, but the more information Fakebox is given, the more useful the results will be.

• Title - Titles can be clickbait or biased
• Content - The textual content of an article can be analyzed to determine if its written like real news or not
• Domain name - Some domains are known for hosting certain types of content, Fakebox knows about the most popular sites

Get started

The best way to get started with Fakebox is to spin up a local instance and play with it.

• Sign into your account and obtain your box key
• Open a terminal and run: docker run -p 8080:8080 -e "MB_KEY=$MB_KEY" machinebox/fakebox
• Go to http://localhost:8080/ in your browser to access the Fakebox console